TIIP Newsletter

Is R&D productivity rising or falling? How valuable is a firm’s portfolio of patents? To answer such questions, scholars and analysts face the daunting task of sorting the wheat from the chaff. It is clear that inventions vary widely in value. Some, like Cohen and Boyer’s gene-splicing technique, will generate lucrative returns over the lifetime of a patent. Most, however, will not: the technology will become obsolete; the patent will be easy to invent around; potential markets will fail to materialize.

Even if it is not possible to predict with any degree of certainty which inventions will be blockbusters and which ones will fail, can we at least improve our forecasting abilities? According to this study by Professors Lanjouw and Schankerman, we can.

In this paper, the authors construct and test a new “composite index” of patent quality. The index is based on both forward-looking indicators—for example, how many times a patent is referenced as “prior art” in patent applications for subsequent inventions—as well as information that is available earlier in a patent’s life, such as the number of claims listed on the patent, the number of references to previous inventions, and the number of countries in which an applicant seeks protection. Although the authors point out that their patent quality index does not perfectly predict the value of an invention, they assert that the two are nonetheless related and examine the issue empirically.

To test how well various indicators predict patent value both individually and collectively the authors use a remarkable database. This database combines detailed information on patents issued in the United States between 1960 and 1991 with information about which patents were renewed (by age 4) or involved in legal disputes. As the authors have demonstrated in prior research, an owner’s decision to renew a patent or to enforce it in court is an indicator of an invention’s private value. An invention’s private value to the firm differs from its social value: on the one hand, consumers may realize benefits in excess of the price of an invention; on the other hand, firms may profit from patents by blocking competitors—a practice without a clear social benefit. The authors use this specific notion of value to assess the statistical predictors in their index. Their analysis is based on a sample of about 8,000 patents in four industries: pharmaceuticals, chemicals, electronics, and mechanics.

The paper is quite comprehensive and contains multiple layers of analysis. Key findings include the following:

1. Four indicators commonly used to make early predictions about the value and technological importance of patented inventions—the number of patent claims, the number times other patents subsequently cite this patent (forward citations), the number of other patents this one cites (backward citations), and the number of countries in which a patent is obtained (family size) —are closely related. Among these, forward citations and claims appear to be the most reliable—that is, the least “noisy—indicators of underlying patent quality.

2. A composite index of quality based on multiple indicators offers additional information. The composite index improves the prediction of expected patent quality. It also predicts fairly well whether owners will decide to renew their patents or take a patent dispute to court.

3. Within the composite quality index, the relative contribution of indicators varies depending on the industry in question. Forward and backward citations are important indicators of patent quality in pharmaceuticals, for example, while family size is considerably more important for electronic and mechanical patents.

4. With the exception of pharmaceuticals, increasing the citation span beyond five years does not improve the quality of information that the forward citation indicator provides. This finding suggests that it is unnecessary to use long citation spans—say, ten or fifteen years—to measure initial expectations about the quality of a patent.

The final part of the paper uses the composite index of patent quality to reexamine an important empirical question: did R&D productivity in the United States decline during the 1980s? The results are somewhat ambiguous. Adjusting for changes in the composite quality index removes much of the apparent decline in research productivity identified in previous studies. Nonetheless, it is unclear whether the authors are controlling for changes in the true underlying quality of innovation or simply capturing unrelated changes in the propensity of firms to cite over time (a key variable embedded in the index). Overcoming such problems of interpretation is clearly an important challenge for future research.


		The Quality of Ideas: Measuring Innovation with Multiple Indicators by Jean O. Lanjouw (Brookings) and Mark Schankerman (London School of Economics) NBER Working Paper #7345 (September 1999)
		--Summary by Rosemarie Ziedonis (Michigan)
		Is R&D productivity rising or falling? How valuable is a firm’s portfolio of patents? To answer such questions, scholars and analysts face the daunting task of sorting the wheat from the chaff. It is clear that inventions vary widely in value. Some, like Cohen and Boyer’s gene-splicing technique, will generate lucrative returns over the lifetime of a patent. Most, however, will not: the technology will become obsolete; the patent will be easy to invent around; potential markets will fail to materialize. Even if it is not possible to predict with any degree of certainty which inventions will be blockbusters and which ones will fail, can we at least improve our forecasting abilities? According to this study by Professors Lanjouw and Schankerman, we can. In this paper, the authors construct and test a new “composite index” of patent quality. The index is based on both forward-looking indicators—for example, how many times a patent is referenced as “prior art” in patent applications for subsequent inventions—as well as information that is available earlier in a patent’s life, such as the number of claims listed on the patent, the number of references to previous inventions, and the number of countries in which an applicant seeks protection. Although the authors point out that their patent quality index does not perfectly predict the value of an invention, they assert that the two are nonetheless related and examine the issue empirically. To test how well various indicators predict patent value both individually and collectively the authors use a remarkable database. This database combines detailed information on patents issued in the United States between 1960 and 1991 with information about which patents were renewed (by age 4) or involved in legal disputes. As the authors have demonstrated in prior research, an owner’s decision to renew a patent or to enforce it in court is an indicator of an invention’s private value. An invention’s private value to the firm differs from its social value: on the one hand, consumers may realize benefits in excess of the price of an invention; on the other hand, firms may profit from patents by blocking competitors—a practice without a clear social benefit. The authors use this specific notion of value to assess the statistical predictors in their index. Their analysis is based on a sample of about 8,000 patents in four industries: pharmaceuticals, chemicals, electronics, and mechanics. The paper is quite comprehensive and contains multiple layers of analysis. Key findings include the following: 1. Four indicators commonly used to make early predictions about the value and technological importance of patented inventions—the number of patent claims, the number times other patents subsequently cite this patent (forward citations), the number of other patents this one cites (backward citations), and the number of countries in which a patent is obtained (family size) —are closely related. Among these, forward citations and claims appear to be the most reliable—that is, the least “noisy—indicators of underlying patent quality. 2. A composite index of quality based on multiple indicators offers additional information. The composite index improves the prediction of expected patent quality. It also predicts fairly well whether owners will decide to renew their patents or take a patent dispute to court. 3. Within the composite quality index, the relative contribution of indicators varies depending on the industry in question. Forward and backward citations are important indicators of patent quality in pharmaceuticals, for example, while family size is considerably more important for electronic and mechanical patents. 4. With the exception of pharmaceuticals, increasing the citation span beyond five years does not improve the quality of information that the forward citation indicator provides. This finding suggests that it is unnecessary to use long citation spans—say, ten or fifteen years—to measure initial expectations about the quality of a patent. The final part of the paper uses the composite index of patent quality to reexamine an important empirical question: did R&D productivity in the United States decline during the 1980s? The results are somewhat ambiguous. Adjusting for changes in the composite quality index removes much of the apparent decline in research productivity identified in previous studies. Nonetheless, it is unclear whether the authors are controlling for changes in the true underlying quality of innovation or simply capturing unrelated changes in the propensity of firms to cite over time (a key variable embedded in the index). Overcoming such problems of interpretation is clearly an important challenge for future research. © 2003. Verbatim copying and distribution of this entire article for noncommerical use are permitted provided this notice is preserved.