At the recent American Law and Economics Association (ALEA) meeting at Columbia, I provided some comments on Intellectual Property Contracts: Theory and Evidence from Screenplay Sales by Milton Harris, Abraham Ravid, Ronald Sverdlove, and Suman Basuroy. The paper works with a fascinating dataset: 1269 contracts for screenplay sales between 1997 and 2003, which are either for a fixed price (averaging $958,000) or contingent on production—but not success—of the script (with average initial payments of $458,000 and average total compensation of $914,000).
Why are many of these contracts contingent when the buyers are typically better at bearing risk? And why are contingent contracts more likely for less experienced sellers? One typical explanation is moral hazard, but apparently that is not an issue in screenplay sales: studios use other writers to edit scripts prior to production, so no further effort from the seller is required once the script is sold. Instead, this paper develops a model in which the seller's competence is not directly observable by either party, and the seller is more optimistic about her competence than the buyer. And as sellers become more experienced, more information is available, which narrows the difference of opinion between the buyer and seller.
I had two general sets of questions about the paper:
First, is this in fact the best model to fit the screenplay data? I think the key feature of the model is the difference in assessment between the buyer and the seller, and the paper emphasizes that the seller's competence is not directly observable by either party. But as acknowledged in a footnote, one could get a similar result from adverse selection and signaling in a world where the seller does have good information, and I'm not sure how to distinguish these stories. Alternatively, one could see similar results with a model in which the parties have the same average valuation, but their risk preferences vary. For example, this paper assumes the buyer has all the bargaining power, and buyers might think there is a much higher variance in results for inexperienced sellers and might thus have a stronger preference in those cases for contingency payments in terms of their own business models.
Second, how well does this model generalize beyond movie script sales to other areas of IP? The paper is framed as a general model of contracts for the sale of IP, so even if this is the best model to explain the screenplay sales data, it is worth thinking about whether other IP transactions generally will look similar.
The description of the writer trying to sell her screenplay sounds like a classic example of Arrow's Information Paradox, in that the buyer can't properly value the product without seeing it, but once he has the information he doesn't need to buy it. IP is usually viewed as a solution to Arrow's problem, but I gather that copyright law doesn't really protect the writers very well because of the idea/expression dichotomy: the valuable part of a screenplay is the idea of the script rather than the particular expression (which is typically completely rewritten). But in that sense, contracts over screenplays are a special case because the parties aren't really contracting over the IP.
Also, the paper states that ignoring moral hazard in IP contracting is justified because "[o]nce a patent or a screenplay is sold, generally no further effort on the part of the seller is required." This may be true for many contracts, but certainly not all. The majority of publicly filed patent licenses include the transfer of know-how in addition to the patent asset, and in the market for film adaptations of books there are at least some prominent examples of authors who were involved in film production and marketing.
I thought this paper was interesting in light of the paper I blogged about recently by Steve Haber and Seth Werfel, which found in a controlled laboratory experiment that inventors were much less likely to accept a lump sum payment for a hypothetical patent if they had the option of getting a contingent payment instead. That paper focused on the role of patent trolls in helping inventors monetize patents, but I'd be curious to know whether typical producers of creative works versus patentable inventions have similar preferences.
Doing an experiment on screenplay writers is surely beyond the scope of this paper, but it might be interesting at least to see if it is possible to add some patent-related dataset to this paper, especially if the paper will continue to be pitched as a general model for IP contracting. From talking with people who know more about patent contracts than I do, my sense is that whether there are contingency payments is largely driven by the business models of the parties, so if there were evidence that the experience of the seller is an important factor, that would be interesting.