Old Method, New Applications in Damages Calculations

In patent damages calculations, an expert generally faces the task of comparing the actual outcome (with infringement) to the outcome in a world without infringement. As stated by the U.S. Court of Appeals for the Federal Circuit:

[T]he statutory measure of “damages” is “the difference between [the patent owner’s] pecuniary condition after the infringement, and what his condition would have been if the infringement had not occurred….” The “but for” inquiry therefore requires a reconstruction of the market, as it would have developed absent the infringing product, to determine what the patentee “would . . . have made.”

Reconstructing the market, by definition a hypothetical enterprise, requires the patentee to project economic results that did not occur. To prevent the hypothetical from lapsing into pure speculation, this court requires sound economic proof of the nature of the market and likely outcomes with infringement factored out of the economic picture.

In economic terms, such reconstruction is called “counterfactual analysis.” While economic methods often “project economic results that did not occur,” and such exercises are widely employed in the economic literature, they are still relatively rare in the analysis of damages, where heuristic and other ad hoc methods often substitute for formal economic methods.

One powerful tool to perform a counterfactual analysis is to construct a model of consumer demand in a market consisting of patented and infringing products. Using the results from such “structural demand estimation,” one can answer questions such as:

• If the infringing product B did not exist, would consumers be more likely to switch to patented product A, or to third-party product C?
• What would be the unit sales of product A if product B did not exist?
• What would be the price of product A if product B did not exist?
• If infringing product B had not implemented the patented feature, how much profit would it have lost, and how much profit would products A and C have gained?
• How would consumers benefit or lose if product B had not been introduced?

Structural demand estimation is a powerful analytical tool that has been used often to construct or reconstruct hypothetical market outcomes (e.g., in the analysis of mergers). It does come at the cost of computational complexity and intensive data usage:  price, unit sales, characteristics of the patented, infringing and third-party products in the same market, in multiple regions and/or time periods. As with merger analyses, which also rely heavily on structural demand estimation, these data are often available in the public domain through data vendors or market reports.

This article presents examples of structural demand estimation applications from the author’s recent damages work.

Example 1

In a pharmaceutical context, the biosimilar defendant is accused of infringing a method patent used in the process of manufacturing a product that competes with the plaintiff’s reference product. The plaintiff claims that non-infringing alternatives are not available. The damages experts are tasked with computing the profit lost by the plaintiff.

The lost profit inquiry asks, had the infringing biosimilar been not available, how would the patentee’s sales, price and profit have changed? A common ad hoc technique is to compute the patentee’s but-for sales by proportionally allocating the sales of the infringing product between the patentee and third parties, based on their current market share. This technique was first approved in State Industries v. Mor-Flo, as a generalization of the so-called Panduit test. Economists refer to this method as “proportional diversion” (since it assumes that the distribution of second choices among consumers is the same as the observable distribution of first choices). This method may yield acceptable results in so-called “homogeneous product markets,” where product features are substantially identical and all consumers have the same preferences for them.

But the accuracy of this method suffers in “differentiated product markets,” in which features differ across product types and preferences differ across consumers. In such cases, there may be good reasons, based on the customer’s revealed preference for the infringer’s price-feature combination, to believe that the patentee’s product is more (or less) likely to have been the second choice of the infringer’s customers, than would otherwise be indicated by market shares alone. For example, a consumer who chooses a Ford pickup truck is disproportionately likely to have chosen another brand of pickup truck, had the Ford not been available, rather than to have chosen a Toyota sedan, even though the sedan has a larger share of the automobile market.

In addition, the proportional diversion assumption does not take into account the effects of price changes due to reduced competition or more complex substitution patterns in differentiated product markets.

Biologic pharmaceutical and small-molecule markets are particularly susceptible to this confusion. In small-molecule markets, generic entrants are medically identical to the incumbent brand (excepting combinations of other ingredients or features). These are thus largely homogeneous product markets except for the branded/generic distinction since certain patients and doctors prefer and are willing to pay a premium for the branded product. This means that, even though generics are virtually homogeneous among themselves, the branded drug occupies a different market niche.

This distinction applies with even greater force to the reference product and its biosimilars, which are not homogeneous among themselves. Having already selected the infringing biosimilar over the reference product, the biosimilar customer may not have chosen the reference product if the biosimilar had not been available, instead choosing another biosimilar. Hence, the Mor-Flo / proportional diversion assumption may be inappropriate and may yield erroneous results.

Structural demand estimation identifies empirically the next-best alternative to each product, based on observed substitution patterns. It also calculates the plaintiff’s profit if an infringing biosimilar did not exist, taking into account the substitution patterns among other competitive products, as well as the recalibration of prices in the counterfactual equilibrium. From these revised prices and quantities, the damages expert can properly measure the but-for profits that the plaintiff would have earned “with infringement factored out of the economic picture.”

Example 2

A smartphone manufacturer is accused of infringing a patent essential to the 5G standard. To determine a reasonable royalty, absent any comparable license, the damages experts must estimate the value of the patented “feature” (here, 5G capability) to the infringer, then apply appropriate apportionment methods to account for the patentee’s share of the value of the standard.

A common approach is the top-down analysis, which involves: (1) determining the aggregate value of the standardized technology (the “top”); and (2) dividing that top among all holders of standard-essential patents, which together create the “value of the feature” (the “bottom”). To determine the top, various methods have been used in the past. For example:

• Some analysts calculate the average price difference between phones with and without the new standard. This approach does not account for other differences that are unrelated to the new technology. For example, 5G prices could be higher because, on average, 5G phones also have more RAM or faster processors.
• For 4G and 3G, some have relied on certain innovators’ so-called “pledge to the market” regarding the aggregate industry royalty burden. Various flaws make this “pledge” far from a good estimate: (1) the “pledge” of certain parties lacked the power to bind other SEP holders; (2) the “pledge” is, at most, a prediction made (3) at a time well before the finalization of the standard, when (4) most of the patents that would eventually be disclosed as potentially essential to the future standard had not yet issued, and many were not even public, nor were (5) the terms on which their owners proposed to license them.
• A few others have implemented a so-called “hedonic regression analysis.”

So far, this last method has proved to be the most economically rigorous and reliable way to estimate the “top,” because it is based on actual transactions that produce “fair market values.” It also accounts for other important differences between smartphones with and without the new standard. However, there are a few shortcomings to this approach. First, it only looks at changes in price associated with the new technology but says nothing about changes in the quantities sold. Second, it only answers the actual question of what consumers paid for 5G but does not consider the dynamic changes to the infringer’s sales had it not implemented 5G.

In the standards context, the standard implementer generally does not have a choice to implement an individual standardized patent; that selection is made by the standard development organization. The implementer’s only choice is whether or not to implement the standard as a whole. Hence, the ultimate question is, if the implementer had not implemented the 5G standard in its products, but its competitors did, what would the implementer’s profit have been?  The profit difference between this hypothetical scenario and the actual world represents the incremental value of the standard to the individual implementer.

Structural demand estimation reliably answers this counterfactual market equilibrium question, taking into account both the higher price and the higher quantities caused by implementing the new standard instead of being limited to the old. While that aggregate increase in profit must of course be subject to further apportionment among individual essential patents for damages purposes, this method reliably captures the incremental value that implementers receive from the decision to implement the standard.

Takeaway

The structural demand estimation method has been successfully applied as discussed above to determine the proper legal counterfactual for intellectual property damages: market equilibrium with “infringement factored out of the economic picture.” Substitution patterns among differentiated products reveal that the value of patented technology may be much greater (or much less) than a patentee or an infringer claims. They also reveal alternative explanations for the observed phenomena: lost sales and price erosion caused by an ostensibly “non-competitive” product, or large consumer willingness to pay for ostensibly non-differentiating technology.

These are just two examples of how structural demand estimation complies with increasingly stringent legal standards, and offers valuable insights that ad hoc, non-structural methods simply cannot provide.