A number of recent papers (particularly those by Helpman and those by Grossman) have considered qualitatively the effects of international IP harmonization following the Uruguay Round TRIPS agreement. In general, they do not support the idea that such harmonization is welfare enhancing. Thurk, a current job candidate who recently signed with Notre Dame, attempts to solve quantitatively for such welfare effects by modeling IPR protection as a discrete-time dynamic game. Firms not on the technology frontier learn from the existing body of knowledge, high IPR encourages more R&D which also increases firm productivity stochastically, and firms trade products internationally. Clearly maximal IP will not be optimal: if IP is high in other countries, I will have low IP to allow my firms to learn rapidly from foreign innovation. The model gives the standard qualitative result that less developed countries, whose firms are further from the technology frontier, prefer less IP protection. Some fancy computational work allows the exact welfare effects to be found after assuming a quasilinear utility function for representative agents. A global welfare maximizer, when a single level of IP protection must be chosen, will lower IP protection by 36% vis-a-vis the US’ current level, and even so such a standard will result in a net welfare loss for less developed countries.
The paper should be seen in some ways as more a computational exercise than anything. I don’t really buy 1) welfare measurements using an arbitrary social welfare function (hence the reason previous work generally just signs the effects), 2) the meaning of numbers like “36% lower IP protection” since the model is discussing utility patents (i.e., IP that goes into production) and TRIPS and the IP index used for data in the paper do not solely relate to that type of protection, and 3) the most common measure in the paper of welfare is average effect, in percent, over countries. Who cares about such an average? I could buy average of people, measured in dollars (i.e., rep agents with quasilinear utility). This materially affects the data, since essentially the paper is treating capital-W Welfare of Denmark as equivalent to Welfare of Brazil, even though Brazil is much more populated. It seems a strange choice.