Topics Contribution of Science to Drug Discovery in Japan, the U.S. and Europe: Implications from Matched Data on Patents and Publications

Printable PDF

Sadao Nagaoka, Professor Emeritus, Hitotsubashi University, Research Advisor, Pharmaceutical Industry Policy Institute
Junichi Nishimura, Professor, Gakushuin University, Visiting Fellow, Pharmaceutical Industry Policy Institute

SUMMARY

1. Introduction

Japan's pharmaceutical industry has been engaged in innovative drug discovery that is highly regarded around the world, but in recent years, its drug discovery capabilities have been declining compared to those of other industrialized countries. As reported by Morimoto (2024) in this issue of Policy Research Institute News, in the first half of the 2000s (2003), 12 of the top 100 products in terms of sales were discovered by Japanese companies, following the United States and the United Kingdom, but by 2022 this number had decreased to 7. During this period, the number of top products by the U.S., Germany, and Denmark expanded, while that by the U.K. declined temporarily but has been expanding in recent years.

The purpose of this paper is to contribute to the analysis of the factors by focusing on the ability of companies to use science. To this end, we constructed and analyzed a database of scientific papers used in the invention process of drug discovery (hereafter referred to as "papers") in each country, matched with pharmaceutical patent data.

The structure of this paper is as follows: Section 2 describes the two datasets (Datasets A and B) constructed for the analysis; Sections 3 and 4 use Dataset A to compare the fields and geographical spread of science utilized in pharmaceutical inventions in the U.S., Europe, and Japan; and Sections 4 and 5 describe the lag in the utilization of science in inventions in the U.S., Europe, and Japan. We also compare the lags between Japan, the U.S., and Europe when science is utilized in inventions, and analyze the effect of having science and inventions created in the same country. Next, Section 5 reports the results of the analysis of the important impact of the use of science on the performance of drug discovery and the differences between Japan, the U.S., and Europe in the use of science, including trends in Korea, Taiwan, and China, using Data Set B.

Survey Methodology

Two datasets were constructed in this paper. First, for drugs approved in Japan by September 2019, we identified U.S. registered patents that correspond to Japanese patents for the drug, and constructed a data set of articles that cite the U.S. patents and are published in the Web of Science (WOS) (Data Set A). The San-Ei Report was used to identify patents protecting each drug, including the patent's classification (i.e., whether it is a substance patent or not). In addition, PATSTAT published by the European Patent Office was used to identify U.S. patents corresponding to Japanese patents. By limiting the analysis to drugs for which U.S. patents existed, we limited the analysis to inventions and drugs with international characteristics. This dataset A also identifies the institutional affiliation of the authors of the papers, which identifies the source country of the papers on which each country's drug inventions rely.

Second, for U.S. patents registered in the U.S. that are classified as drug inventions out of 35 technical fields (field 16) based on the IPC and Technology Concordance of the World Intellectual Property Organization (WIPO), PATSTAT of the European Patent Office and Professor Marx of Cornell University The matched data of patents and papers are constructed using the PCS data published by PATSTAT of the European Patent Office and Prof. Marx of Cornell University (Dataset B). Dataset B includes many patents in the field of pharmaceuticals that did not eventually reach the market as drugs, and it also covers recent inventions, allowing for analysis of more recent trends. Both sets of data allow for complementary analysis.

Results

(1) Geographical distribution of science used for drug discovery and the spread of the field

(i) Explanation of data

Of the 2,288 Japanese patents for pharmaceuticals approved in Japan through September 2019, there are 1,119 Japanese patents with U.S. patents that cite articles published in WOS, and there are 582 by number of active ingredients. In this paper, 134 ingredients are covered in Dataset A as a result of selecting the estimation sample under the conditions explained below. The earliest approval of a drug with an invention meeting these conditions was in October 1991.

In this paper, the papers on which the drug discovery relies are identified by citations in U.S. patents. When applying for a patent in the U.S., the applicant is required to file an Information Disclosure Statement (IDS) with the Patent Office, which lists prior art references of which the applicant is aware. The U.S. Patent Office examiner checks the contents of the statement and supplements the prior art references with his/her own literature search, the results of which are published as a list of prior art references for U.S. patents. The WOS papers that exist in the list are the subject of our analysis in Dataset A. For the papers, we obtained information on the publication date and the affiliations of the authors. We calculated the difference between the publication date of the paper and the priority claim date of the patent, and calculated the citation lag from the paper to the invention. Information on the location of the institutions to which the authors belong was collected from web information using CHAT-GPT, and was used to determine the geographical distribution of science.

In the following analysis using Dataset A, we focus on substance patents, which are considered to be highly important in drug discovery innovation, and identify corresponding U.S. patents and the papers they cite. The descriptive statistics from the analysis sample for each applicant country are shown in Table 1 in the Appendix. It can be seen that the number of papers cited per patent is 2.8 times higher for the substance patents of U.S. firms than for Japanese firms.

Although the affiliations of the authors of the papers can be divided into universities, public research institutes, and companies, most of the papers that are frequently cited by patents are from universities and public research institutes. The top 10 most frequently cited institutions by material patents in our sample were universities (e.g., University of Cambridge, University of London, University of Sheffield, MIT) or public research institutes (e.g., MRC, Scripps Research Institute, NIH), with the exception of one company.

When an author's institutional affiliation spanned multiple countries in a paper, author-institution dummies were created for each major country in each paper to reflect the contribution of each country. For example, in the analysis of citation lag from papers to patents, if a paper is a joint paper between a U.S. and Japanese university, the dummy for the country where the author's institution in Japan and the U.S. is located takes 1 and is reflected in the estimation. The same is true for joint patent applications.

(2) Geographical distribution of science used to generate substance patents

In the following analysis, we focus on patent-paper combinations that have a substance patent for a drug launched in Japan, a WOS citation from the corresponding first registered patent in the U.S., and a non-negative citation lag since it was issued after 1980 (see Table 1 in the Appendix for descriptive statistics; papers with negative citation lag are inventions and may have been added later and are not included in the following analysis. (See Appendix for a discussion of why the first U.S. patent was included.)

Table 1 shows the country of residence of the applicant of each patent and the country distribution of the institution to which the author of the WOS paper cited by that patent belongs, according to this dataset. The horizontal axis is the country of residence of the applicant of the patent, and the vertical axis is the country of residence of the institution to which the author of the paper belongs. In this table, for ease of understanding the correspondence with the papers, the author countries or applicant countries are aggregated to one country when there are multiple countries (the lower-ranking country in terms of the number of patents and papers is given priority in assigning papers and patents; however, in Japan and the U.S., the U.S. is given priority). Table 2 in the Appendix shows the case in which the author and applicant countries are not aggregated (aggregation by so-called Whole Counts), and the trends are very similar (the ratios of each type are also close). Thus, the basic findings from Table 1, discussed below, are not affected.

The total number of papers cited for drug discovery in each country is 6,680 (the number of papers is weighted by the number of citations; if one paper is cited in two patents, it is counted as two papers). In the table, only six specific countries with a high track record in drug discovery are listed as the applicant's country of origin. The percentage of U.S. applicants citing papers from their own country is about 54%, indicating that foreign papers also play an important role in the process of drug discovery in the U.S. Of the foreign papers, 5.8% of the papers on which U.S. drug discovery relies originated in Japan. Next, Japan cites its own papers 11% of the time, and about 90% of the time, it relies on foreign papers. Germany, the U.K., Switzerland, and Denmark, where the level of drug discovery is high, show the same trend as in Japan. In the U.K., however, the citation rate of home papers is relatively high at 18%, but about 80% of the citations are to foreign papers.

The share of each country in the total frequency of utilizing articles is shown in the bottom column. The share of the U.S. is 56% (3,746/6,680), followed by Japan at 14.4%. The rightmost column of the table shows the distribution of authors by country of institutional affiliation. Scientists in the U.S. supplied 48% of the 6,680 papers, followed by those in the U.K. and Japan with 10% and 7%, respectively.

Figure 1 shows the correlation between the logarithmic values of the frequency with which papers of authors in each country are cited (supply measured by the number of citations of papers) on the horizontal axis and the logarithmic values of the total number of papers cited by applicants in each country (the amount of papers utilized) on the vertical axis. A positive relationship is expected between the supply of papers and the amount of papers utilized in each country, since both increase as the size of the country increases, and both also increase if the research human resources in the life science field expand in countries of the same size.

Table 1 and Figure 1 show that the amount of utilization of papers is more concentrated in specific countries. In the top three countries in utilization and supply in Table 1 (Japan, the U.S., and the U.K.), the total utilization share is 77%, and the total supply share is 65%. Countries such as Belgium and Denmark, which have a small share in the supply of papers, have a high share in drug discovery. This suggests that building research capacity, mainly in universities, may be quite different from building drug discovery capacity in companies, which require the ability to conduct long-term projects and raise risk funds until commercialization.

(iii) Dominance of home country inventors in science utilization

To leverage science for invention, knowledge (know-how) embodied in the human capital of researchers who have made scientific inventions is important (Zucker et al. (2000)). To obtain knowledge transfer, at the dawn of the biotech industry in the U.S., biotech firms located in the vicinity of universities where the core inventors were affiliated (Kenney and Mowery (2014)). In addition to knowledge transfer, geographic proximity makes it easier to conduct industry-academia collaborative research and exchange at academic conferences, and furthermore, industry-academia collaborative research in the home country is more likely to receive financial support from the home country. These are the advantages of homegrown inventors in the utilization of homegrown science.

In order to get an idea of this advantage, Table 2 compares the probability of home-country firms using home-country science (for example, Japanese firms using Japanese science) and foreign firms using home-country science (U.S. firms using Japanese science), using the data in Table 1. The patent data used here are all connected to drug discovery (market launch), and if the home country firm has an advantage, the probability of the former is expected to be higher.

In fact, in the six top drug discovery countries (Japan, the U.S., Germany, the U.K., Switzerland, and Denmark), the former is considerably larger than the latter, with the exception of Switzerland. For example, in the case of Japan, the probability that a foreign company will use Japanese science is 6%, while the probability that a home country company will use its own science is 11%. Similarly, in the UK and Denmark, the probability of home country firms using home country science is high. In the U.S., the probability of using U.S. science by foreign firms is 40%, and the probability of using U.S. science by U.S. firms is 1.4 times higher, at 54%.

On the other hand, in the Netherlands and Spain, there are no drug discoveries (drugs with substance patents that have been launched in the Japanese market) by the home country firms, and only foreign firms use home country science.

(4) Areas of science used for drug discovery

Table 3 shows the fields of science contributing to drug discovery by the applicant's country of residence. The 22 categories of journals published in WOS (Essential science indicators) are used as the fields of science.

The number of fields is counted if there is at least one paper in the field. As shown in the total column, almost all of the substance patents for pharmaceuticals launched in Japan are based on papers in all fields (only in the field of economics and management, there are no patents that are based on papers in all fields). This indicates that a wide range of fields of science is utilized in drug discovery. The U.S. leads in the utilization of a wide variety of sciences, utilizing the largest number of sciences in 21 fields. Japan, Germany, and the U.K. were half of that number (11 fields). (Note that the total number of citations in Table 3 does not match that in Table 1 because the field classification of some papers in the WOS database is not clear.)

The lower part of Table 3 shows the results after further disaggregating the fields into four major fields (8 fields for basic life, 2 fields for chemistry and pharmacology, 2 fields for clinical medicine, and 10 fields for ^othersⅰ). As indicated by the total on the far right, these 12 fields account for the majority of citations, with basic life sciences accounting for 53%, chemistry/pharmaceutical sciences 24%, and clinical medicine 22%. There are differences in the science relied on in each country. For example, in Germany, the share of articles in chemistry and pharmacology is as high as 41%, and the share of clinical medicine is also high at 33%. In Denmark, by contrast, basic life accounts for a high 63%. The U.S., Japan, the U.K., and Switzerland are similar in the areas of science that they utilize.

(2) Lag from science to drug discovery

The short lag time between the utilization of scientific advances in drug discovery enables innovations to be realized earlier and enriches society. In addition, from the viewpoint of competition in drug discovery, early utilization of science is considered to be an important factor. In this section, we surveyed the substance patents that were the first to cite the relevant paper in terms of the ^paperii), and calculated the difference (citation lag) between the publication date of the relevant paper and the filing date (priority claim date) of the substance patent. In the following, we use the citation lag as the explained variable to analyze how the lag differs among Japanese, U.S., and European firms after controlling for the field of the article relied upon, whether the applicant and the author of the article are located in the same country, and these factors.

Distribution of lag time to drug discovery based on survival time analysis

Before conducting the regression analysis, we show in Figure 2 how the citation lag to drug discovery differs among Japanese, U.S., and U.K. firms, among others. The vertical axis of this figure shows the "survival probability" (i.e., the probability that at each elapsed time point, the drug has remained only in the paper and has not yet been utilized in a substance patent to be launched in the market). The horizontal axis is the lag between the year of publication of the paper and the year of filing of the invention, and as it gets longer, the probability of not being utilized in drug discovery decreases (the probability of being utilized in drug discovery increases). In the dataset A used here, only papers that have been utilized in drug discovery are included in the analysis, so the probability starts from 1 and becomes nearly 0 after about 20 years.

Figure 2 shows that the blue line for Japan (JP) is at the top, while the line for the United Kingdom (GB) is at the bottom, and the United States is in the middle. It can be seen that it takes about 7 years in the UK, 9 years in the US, and 11 years in Japan to reach the 75% probability that a paper will be used for drug discovery. The difference in the lines for these three countries is statistically significant (we will verify by estimation whether it is significant after controlling for other factors). As shown in Table 1 in the Appendix, Japanese firms utilize only a relatively small number of papers per substance patent in their inventions compared to firms in the U.S. and the U.K. The lag results also indicate that they are slow in taking advantage of scientific advances. It is expected that Japanese firms will be able to enhance their achievements by seeking out promising research opportunities earlier and conducting research faster to take advantage of them, and this ability will be further demanded of Japanese firms.

(2) Analysis of lag to drug discovery by regression analysis

In the following, we use regression analysis (Cox proportional hazards model) to evaluate whether differences in patent applicant countries ("U.S." basis) and whether the applicant organization and the article author organization being in the same country significantly predict citation lag to drug discovery. The model estimates the probability of a patent application event leading to drug discovery (i.e., a substance patent citing a paper) assuming that exogenous factors (explanatory variables) have a proportional impact through time. We control for the country of origin of the paper author's institution (the "other country" criterion), the 22 fields of the paper, the importance of the paper in science as measured by its citation in WOS, and the year of publication of the paper. Since each substance patent relies on multiple papers, the errors from these papers are correlated, and we control for this through clustering (e.g., a more accelerated patenting procedure in the review process would lead to a shorter lag from all papers cited by this patent). There were 3,602 papers analyzed, of which 1,776 were by U.S.-located authors, and U.S. companies were the first to utilize (cite) 2,257 papers, including foreign papers, in drug discovery. The number of papers with Japanese authors is 231, and Japanese companies are the first to utilize 415 papers in drug discovery.

Table 4 shows the results. The table shows that the coefficient for Japan (on a "U.S." basis) is negative, indicating that patent application events leading to drug discovery are less likely to occur earlier in Japan than in the U.S., but this is not significant. On the other hand, the coefficients are positive for all European countries relative to the U.S. and significant for the U.K., Switzerland, and Denmark. The difference in coefficients between these countries and Japan is also larger and more significant in Europe, indicating that patent application events occur earlier in these European countries than in Japan. This result is consistent with Figure 2.

Next, looking at the coefficients for the author's country (based on the "other countries" criterion), their magnitude is smaller than that of the coefficients for the filing country, and the same is true for the difference with the United States. For example, the difference between Japan and the U.S. (Japan-U.S.) is -0.34 for the filing country and 0.14 for the author country. The difference between the U.S. and Germany is 0.48 for the filing country and 0.06 for the author country. This indicates that the ability of the users (corporate applicants) is a more important determinant of the size of the lag.

As for the control variables, the number of citations of a paper is not significant. It is likely that higher quality papers can lead to better inventions, but the more innovative the invention, the longer it may take. Finally, the fact that the applicant and the author institution are located in the same country tends to hasten the drug discovery event, but is not significant.

(3) Use of Science and Drug Discovery Performance

The analysis thus far confirms that there are significant differences in the use of science between Japan, the U.S., and Europe. In the following, we analyze the impact of these differences on drug discovery performance. In the following analysis, we will use Dataset B to utilize U.S. patents registered until recently, and also check the situation in more recent years. In addition, pharmaceutical patents that did not result in drug discovery (marketed products) will be included in the analysis, allowing for a more comprehensive data set to examine the extent to which science has an impact on drug discovery performance.

(1) Explanation of data and estimation model

Using the European Patent Office's PATSTAT, data on U.S.-registered patents whose applications were filed in 1980 or later are included in the analysis. The pharmaceutical patents that have been launched in the Japanese market analyzed in Dataset A are categorized under the WIPO 35 technology classifications of "Organic Fine Chemistry" (Classification 14), "Biotechnology" (Classification 15), and "Pharmaceuticals" (Classification 16). The unit of analysis is the patent unit. The unit of analysis is the patent unit, but we focus on the earliest filed patents in PATSTAT's patent families.

As an indicator of the performance of drug discovery innovation, we use whether the invention in question is highly cited by subsequent patents (top 5% of citations) and whether it has become a drug patent that is marketed in Japan ⁽ⁱⁱⁱ⁾. As explanatory variables, the main variables are the number of papers on which the invention relies (logarithmic number with 1 added) and the minimum citation lag with papers on which the invention relies. In addition to this, we use the logarithmic number of inventors, a dummy for the presence of inventors living abroad, the median backward citation lag from prior patent literature, and the number of citations to inventions by inventors living abroad (logarithmic number plus one). The model assumes that the combination of knowledge is the source of inventions, including science (see Nagaoka (2024 forthcoming) for details of the model's concept and analysis for all industries).

Table 5 shows the trends in the share of pharmaceutical patents in the top 5% of citations for applicants from the six leading industrialized countries in drug discovery and China, South Korea, and Taiwan, three countries that have been increasing their drug discovery power in recent years. In the 1980s, Japan had a 7% share, second only to the U.S., but in the 2010s, that share has decreased to 2.7%, half the share of the U.S. in the 1980s. The U.S. still has a 70% share in the 2010s, although this share is down from 75% in the 1980s. In Europe, the U.K. and Switzerland are increasing their share. China, South Korea, and Taiwan have increased their shares from 0% in the 1980s to 2%, 2%, and 1.4%, respectively, in the 2010s.

As can be seen in Table 6, this can be attributed to the increasing sophistication of inventions in the Western countries, such as the greater use of scientific advances in inventions. Although Japan has increased the number of science utilization per substance patent, in absolute numbers it is significantly inferior to the Western countries, and the gap between these figures has been widening in recent years. In addition to the expansion of R&D, China, South Korea, and Taiwan are also experiencing an increase in the sophistication of inventions through the use of science.

2) Use of science and estimation results of drug discovery performance

Table 7 shows the relationship between the level of science utilization and drug discovery performance. The level of science utilization is measured by two variables: the number of cited papers itself and the logarithm of the number of cited papers plus one. The latter variable, which is logarithmic, reflects the effect of diminishing returns due to overlap between papers. The leftmost result, Model 1, shows how the probability of a patent in the pharmaceutical field being in the top 5% of citations in U.S. patents depends on a number of factors. Model 2 shows the results for all technical fields, including pharmaceuticals, for comparison. Model 3 shows the results for whether or not the invention in question has become a marketed substance patent in the Japanese market. Both models were estimated using linear probability models. In each model, we control for fixed effects of patent application year and application organization, and estimate robust standard errors clustered by organization. Model 2 also introduces WIPO technology field dummies. Since the estimation uses the lag of citations from papers as an explanatory variable, only inventions that cite papers are included in the analysis, but estimation without this variable did not significantly change the main results. Table 2 in the Appendix shows the descriptive statistics for each variable.

According to Model 1 in Table 7, drug discovery performance, as measured by the probability of being in the top 5% cited, is significantly higher when the citation level of the paper is high and the citation lag is small. The number of papers is a significant off-log variable. Based on the estimated coefficient values, the probability of being a top 5% cited invention is 0.9 percentage points higher when evaluated at the average citation level (29 in the original number, which is converted to a logarithm at 3.4) compared to when papers are not cited. Also, if the citation lag of an invention from a paper to an invention is shortened by one year, the probability increases by 0.1 percentage point. The results indicate that there is a fairly large impact on the use of science advances in inventions.

Next, as Model 3 shows, for the probability of a patent of substance becoming a marketed patent of substance, only the logarithm of the paper is significant for the utilization of the results of the paper. As in the case of the probability of being in the top 5% of citations, when evaluated at the average citation level, it is 0.24 percentage points higher than in the case of not being cited. The proportion of marketed substance patents in the sample used in this paper is 0.14%, so the impact is significant. The coefficient of lag from the paper is negative but not significant. Among the marketed products, there is a possibility that the significance is high for drugs with high innovativeness such as new mechanism of action, but this is a subject for future research.

Regarding the other variables, factors other than science, larger inventor team size, shorter lag from prior art literature, and larger level of citations to inventions by foreign-resident inventors all increase drug discovery performance ^ⅳ). As Model 2 shows, the results are consistent with those estimated for all fields, although the magnitude of the coefficient values are different. It can be said that the broad and early application of new knowledge to inventions, including drug discovery, generally enhances the performance of inventions.

4. conclusion

Drug discovery takes advantage of scientific advances in disease mechanisms, targets, mechanisms of action of candidate molecules, structures of candidate molecules, methods of their manufacture, delivery of drugs to their targets, screening methods, etc. Science is advancing rapidly, and the science used for drug discovery is also diversifying. This is also true for small molecule drugs.

As the analysis in this paper shows, the ability to leverage science is an important contributor to influential inventions and the probability that these inventions will be used in pharmaceuticals. At the same time, Japanese companies are less likely than their Western counterparts to take advantage of scientific advances in drug discovery, and when they do, they do so at a slower rate than their Western counterparts. In recent years, the ability to utilize such science has increased in China, Korea, and Taiwan.

Based on the above results, it is important to strengthen the ability of Japanese companies to utilize science. One of the core elements is human resources, especially the development and recruitment of Ph. D. or doctoral dissertation, a significantly higher percentage of inventors recognize that universities and scientific and technical papers are very important sources of knowledge for their inventions (Nagaoka et al., 2012, p. 9). D. in Japanese firms is thought to be about half that of U.S. ^firmsⅳ ), and there is significant room to expand this. In Japan, the existence of doctoral dissertations has played an important role in the advancement of human resources within companies and in industry-academia collaboration, but the fact that this has been greatly reduced in recent years is a major lost opportunity, and its revival is desirable.

In the direction of corporate research, it is also important to strengthen efforts in cutting-edge research from a long-term rather than short-sighted perspective. In the course of such research, human resources will be developed and global knowledge will be acquired.

Advanced research and human resource development by domestic universities are also very important to support such long-term research. In order to strengthen the ability to utilize science, it is desirable to successfully drive both human resource development and R&D strategies.

References

Morimoto, J., 2024, "Nationality of Companies Generating Top Global Sales of Pharmaceuticals: Trends in 2022," Pharmaceutical and Industrial Policy Research Institute, Policy Research Institute News No. 71

Sadao Nagaoka, Naotoshi Tsukada, Koichiro Ohnishi, and Yoichiro Nishimura, 2012, "Innovation Process in Japan in the Early 2000s from the Perspective of Inventors," RIETI Discussion Paper Series 12-J-033

Sadao Nagaoka, 2016, New Drug Creation: Exploring the Sources of Innovative Pharmaceuticals from Japan, Nikkei Business Publications, Inc.

Nagaoka, Sadao (ed.), 2024, "An Examination of the Innovation Capability of Japanese Industry" (forthcoming), University of Tokyo Press

Fleming L., H. Greene, G. Li, M. Marx, D. Yao, 2019, "Government-funded research increasingly fuels innovation," Science 21 . june. vol 364 issue 6446

Kenney Martin and David C. Mowery, 2014, Public Universities and Regional Growth: Insights from the University of California, (ed.) Stanford University

Poege F., Harhoff D., Gaessler F., Baruffaldi S., 2019, "Science quality and the value of inventions," Science Advance, 5 :. eaay7323, 11 December 2019

Zucker, L. G., M. R. Darby and M. Torero [2002]. , "Labor mobility from academe to commerce". Journal of Labor Economics 20(3), 629-660)

Acknowledgments

This research was supported by Grant-in-Aid for Scientific Research, KAKENHI KAKENHI B ("Research on Drug Innovation and Incentives", 18H00854) and by the Institute of Economics and Management, Gakushuin University (Research Project: Diffusion of Drug Innovation and Its Economic Effects). Information on the location of affiliated institutions was collected by Sho Watanabe, Graduate School of Economics, Gakushuin University. We would like to thank the researchers at the Pharmaceutical and Industrial Policy Research Institute (PIIPRI) for their helpful comments on the research in this paper. We would like to take this opportunity to express our deep appreciation.

Appendix 1: Dealing with the possibility of additional paper citations after the fact, such as negative citation lag from paper to invention

One of the findings of this data construction is that the citation lag from papers to inventions is negative in a considerable number of cases. Essentially, the lag should be positive for prior literature. (In the U.S. patent examination process, there is an obligation to provide information on patentability even after the application is filed until registration. (In the U.S. patent examination, there is an obligation to provide information on patentability even after the application is filed until registration.) As clarification of this issue is a subject for future research, negative lags from papers to inventions were excluded from the analysis of lags in Dataset A (the percentage of such lags was used as a control variable in Dataset B).

Another avenue for adding a posteriori citations to articles is through continuation and divisional applications. In the U.S., there are fewer restrictions on this, resulting in 1,676 U.S. patents, 50% more than the number of Japanese patents (1,119). New citations to papers are also added on occasions such as continuation applications for the same invention. These additional cited papers are assumed to be less important as knowledge to the initial invention and are excluded from this analysis.

Return to News List

TOP