Did TRIPs spur innovation? An analysis of patent duration and incentives to innovate.

• Author: Abrams, David S.
• Position: Agreement on Trade-Related Aspects of Intellectual Property Rights

How to structure IP laws in order to maximize social welfare by striking the right balance between incentives to innovate and access to innovation is an empirical question. It is a challenging one to answer, both because innovation is difficult to value and because changes in IP protection are rare. The 1995 TRIPS agreement provides a unique opportunity to explore this question for two reasons. First, the implementation of the agreement was uncertain until shortly before adoption, making it a plausibly exogenous change in patent duration. Second, the nature of the law change meant that the patent-duration change was heterogeneous across patent classes. Using both patent counts and citation-weighted counts, I am able to exploit the TRIPS-induced law change to empirically evaluate the impact of patent duration on innovation. I find evidence for an increase in innovation due to patent-torn extension following TRIPS. Both patent counts and citation-weighted counts increased more after TRIPS in those classes that received greater expected term extensions relative to classes receiving shorter extensions. While the precise calibration of innovation valuation is difficult, this Article provides the first attempt to empirically evaluate its response to a major change in patent duration from the TRIPS agreement.

INTRODUCTION I. BACKGROUND A. The TRIPS Agreement B. Related Literature C. Theoretical Background II. DATA III. ECONOMETRIC METHODOLOGY IV. RESULTS V. DISCUSSION CONCLUSION APPENDIX INTRODUCTION

In 1994, as part of the General Agreement on Tariffs and Trade (GATT) that created the World Trade Organization (WTO), the United States made the largest change in patent terms in over forty years. (1) In order to conform to the Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS), (2) the United States changed the duration of patent protection from seventeen years from grant date to twenty years from application date. (3) This change affords the opportunity to learn about one of the most basic issues in IP: the relationship between the quantity of innovation and the duration of IP protection.

In this Article, I explore this relationship between duration and quantity using U.S. patent application and citation data around the dates of the implementation of the TRIPS agreement. I use a difference-in-difference framework--exploiting the heterogeneous impact of the change across patent classes--based on Patent and Trademark Office (PTO) processing time. I find a statistically significant relationship between the magnitude of the term extension resulting from the TRIPS law change and patent count. Further, I find that this relationship persists when using citation-weighted patent counts as the dependent variable, which is arguably a better proxy for value of innovation. I do not find a significant increase in the average number of citations to patents receiving longer extensions relative to those receiving shorter extensions following the change.

Standard theory argues that an increase in the duration of patent protection has two primary effects: an increased incentive to innovate (created by monopoly profits) and increased deadweight loss (due to exclusive rights). (4) The optimal patent term is that point at which the marginal benefit from increased innovation is exactly offset by the marginal cost of the deadweight loss created by the patent right.

Determining the optimal patent term is extremely important from a policy perspective. If patent protection lasts too long, the monopoly-like deadweight loss, caused by the patent's conferral of exclusive rights, outweighs the additional innovation such rights will spur. On the other hand, a patent term that is too short will yield underproduction of innovation, leading to a decrease in productivity and growth. In order to find the optimum empirically, it is necessary to estimate the elasticity of production of innovation with respect to the duration of protection. Previously, the ability to empirically evaluate the optimal patent length was limited by a dearth of data, a lack of computing power, and an absence of a change to patent length with which to evaluate the elasticity. (5)

This Article works towards evaluating the elasticity of production of innovation by examining data from 1990 through 2000 in light of the 1995 change in patent protection. The change in duration provides the denominator for the elasticity calculation (which I calculate separately by patent class). Though the change was not completely unanticipated, it was part of a global-trade accord that encompassed myriad issues beyond intellectual property. Thus, the law change may be viewed as plausibly exogenous (this assumption will be examined more closely later in the Article) and allows for the identification of the impact of a modification in patent duration.

The numerator in the elasticity calculation should ideally be the percentage change in the total social welfare derived from patented ideas--an exceedingly difficult quantity to measure. In this Article, I take two approaches to approximating this quantity. (6) The first approximation makes use of simple patent counts as a proxy for the value of innovation, examining the number of patents obtained prior to and after the 1995 law change by patent class. While this method should yield the correct sign for the elasticity, the magnitude could be substantially off unless the value of the marginal patentable innovation is constant, which seems highly unlikely. Nevertheless, using patent counts as a measure of innovation has a long history and is informative about the response of innovators to incentives. (7)

Previous research has shown that patents vary widely in value. (8) A closer approximation of total patent value may be obtained by weighting patents according to how many citations they receive from subsequent patents. (9) This method accounts for some of the heterogeneity in patent values and allows for weaker assumptions about the value of the marginal patentable innovation. (10) I make use of the National Bureau of Economic Research (NBER) patent database, which contains patent citations, in order to obtain a better estimate of the numerator. (11)

The rest of the Article proceeds as follows: Part I discusses the background of the TRIPS agreement and the subsequent law change, related literature, and theoretical framework used to analyze its impact. The data are found in Part II and the econometric methodology is explained in Part III. Part IV contains the main results from both the patent-count and citation approaches. Part V discusses the results and their limitations.

1. BACKGROUND

   1. The TRIPS Agreement

      The TRIPS agreement grew out of the Uruguay Round of GATT negotiations, which lasted from 1986 through 1994 and led to the creation of the WTO. (12) TRIPS covered many aspects of intellectual property and required harmonization of IP laws among the developed signatory countries. Within the realm of patents, the key requirements of TRIPS were that patents be made available without discrimination to citizens of TRIPS signatory nations on both products and processes (13) and that the protection extend for a minimum of twenty years. (14)

      At the time of TRIPS's passage, U.S. patent law provided for seventeen years of patent protection, as measured from the patent grant date. Thus, TRIPS necessitated a significant modification of U.S. law. The seventeen years of patent protection in the United States was not derived from an economic calculation, as advocated in this Article. Initially, U.S. law was modeled on English law, which set an initial fourteen-year patent term based on the expected training period for two sets of apprentices. (15) Nordhaus points out that "[a]fter some further compromise it was decided for the United States that 2.43 apprentices, or 17 years, would be the proper length." (16)

      For the purposes of this analysis, there are three relevant dates to consider that could have potentially affected innovative activity, as displayed below in Figure 1. The first is the date that the final package negotiated through the Uruguay Round was signed, April 15, 1994. (17) The second is the date the U.S. Congress ratified the package, December 8, 1994. (18) The third is the date that the change became effective in the United States, June 8, 1995. (19) Based on news reports at the time, it is clear that while the April 15, 1994, signing of the agreement was symbolically important, it was far from certain that the agreement would be effectuated. (20) It was not until ratification by Congress that TRIPS was expected to be implemented.

      [FIGURE 1 OMITTED]

      Difference in difference is an economic technique that may be used to isolate the effect of a law change such as that brought on by the TRIPS agreement. One major application of the technique is to avoid ascribing a causal effect to a law change that is really due to a contemporaneous change or a time trend. This is a significant problem with the single difference, or before-after analysis. With the difference-in-difference approach, one takes the before-after difference for a group that should be affected by the law change and takes a second difference with the before-after difference of a group that should be unaffected (or less affected). This is akin to creating an experimental and a control group.

      For example, consider a law that becomes effective in January 2010 that reduces the cost of bankruptcy but that is phased out for individuals earning above $100,000. A comparison of bankruptcy rates between 2009 and 2010 shows an overall increase of ten percent, but the concern is that this may be due to factors other than the law. Now consider the difference-in-difference approach, comparing the before-after change in bankruptcy rates between those making over $100,000 with those making less. One finds that bankruptcy rates increased by eleven percent for those earning...