John R. Allison

On the Feasibility of Improving Patent Quality One Technology at a Time: The Case of Business Methods

Following the Federal Circuits 1998 decision in State Street Bank & Trust Co. v. Signature Financial Group, Inc. holding that there was no per se exclusion from patentability for software-implemented business methods, the number of applications for and grants of such patents on increased dramatically. A large number of critics castigated them for two primary reasons: First, some contended that patents should simply not be allowed on such subject matter. Second, a much larger number of critics asserted that these patents were of singularly inferior quality because they had been issued without due consideration to the relevant prior art-documentary evidence of what others had done in the past. In response to these criticisms, the United States Patent and Trademark Office (PTO) instituted an initiative in March 2000 that it called the Second Pair of Eyes Review, intended to make it more difficult to obtain business method patents. Among other features, the SPER program called for a second-level review upon allowance for patent applications in Main Class 705. Each of the authors had independently conducted previous research leading us to conclude that software-implemented business method patents cited prior art that was not inferior in quantity or quality to other types of patents, and that the many prior art-related criticisms were not supported by empirical evidence. Although some business method patents undoubtedly were of low quality, one finds low-quality patents in all fields. Be that as it may, we believe it important to empirically assess the effects of the SPER initiative. The program requires substantial resources, and if it has not had significant positive effects, these wasted resources constitute a tax on innovation because the PTO is supported by user fees. Perhaps a more important reason for evaluating SPERs effects, however, is that the PTO declared it to be a success and announced plans to expand it to other fields. Whether or not the agency expands the program in the near term, it is important to know whether targeting patent reforms at particular technology fields is a model that should ever be replicated. Employing negative binomial regression models, we found that the SPER initiative did indeed have significant positive effects on the quantity and quality of prior art cited in patent applications to which it applied. Using logistic regression, we determined that the SPER initiative also led to a significant increase in the number of Main Class 705 patents that cited at least some prior art. Moreover, we found that, after SPER, examiners themselves added significantly more prior art references that did examiners of closely related patents with Secondary 705 classifications. Success within this narrow operative domain, however, is seriously tempered by a number of factors, leading to the conclusion that attempting to improve patent quality one technology at a time is not a model that should be repeated. Because of definitional problems that are impossible to overcome, and because the patent classification system is very poorly suited for this kind of task, such an approach to patent reform is and always will be grossly underinclusive and subject to gaming by both patent applicants and beleaguered patent examiners. Attempts to improve patent quality should not focus on a single field.

The Disputed Quality of Software Patents

We analyze the characteristics of the patents held by firms in the software industry. Unlike prior researchers, we rely on examination of the individual patents to determine which patents involve software inventions. This method of identifying the relevant patents is more laborious than the methods that previous scholars have used, but it produces a dataset from which we can learn more about the role of patents in the software industry. In general, we find that the patents computer technology firms obtain on software inventions have more prior art references, claims, and forward citations than the patents the same firms obtain on non-software inventions. We also find that the patents that software firms obtain on software inventions also have more prior art references, claims, and forward citations than the software patents obtained by the firms that derive revenues from other product lines. Finally, we conclude that the patents of the largest firms are no better (or worse) than the patents of the smallest firms, belying the idea that large firms are plagued by challenges based on the worthless patents of their smaller competitors. The paper closes with a brief discussion of the implications of our empirical analysis. The findings undermine the strongest criticisms about the low quality of software patents. It is simply not accurate to say that software patents as a class have remarkably low numbers of prior art references and forward citations. Thus, they cut against technology-based patent reforms designed to make it more difficult to obtain software patents. On the other hand, the evidence that small firms are no less capable than large firms at producing quality patents vitiates concerns that higher hurdles at the early stage of the patenting process would disadvantage smaller inventors in particular.

Data disclosure and data sharing in scientific research.

Data sharing is examined for its bearing on (i) quality assurance and (ii) extensions of results in scientific research as well as (iii) part of a tradition of openness in science. It is suggested that sharing can be accomplished in a simple manner that is also sufficiently flexible to fit varying individual situations by asking authors of data dependent articles and grant proposals to footnote (a) whether they are willing to make their data available to others and, if so, (b) how the data may be accessed. Appendices report results from a survey of current policies and practices in professional societies and in Federal government fund granting agencies. Emphasis is on the social and management sciences.

Patent Litigation and the Internet

Using both univariate comparisons and multiple regression techniques, we find that: (1) Internet patents and their two subtypes, broad Internet business models and narrower Internet business techniques, were litigated at a far higher rate than other (non-Internet patents, or NIPs) — they were between 7.5 and 9.5 times more likely to end up in infringement litigation, depending on the model we used. (2) Within the category of Internet patents, those on business models were litigated at a significantly higher rate than those on business techniques. (3) Across both Internet patents and NIPs, patents issued to small entities, especially individuals and small businesses, were much more likely to be litigated than those issued to large entities. (4) Patents of all kinds with more independent claims were significantly more likely to be litigated than those with fewer independent claims. (5) Including both Internet patents and NIPs, litigated patents received many more forward citations — citations received from later patents — than did unlitigated patents. (6) Patents issued to foreign entities were significantly less likely to be litigated than patents issued to U.S. entities. (7) The more time that an application for an Internet patent or NIP had spent in the PTO prior to issuance, the more likely it was that the patent granted from that application was to be involved in infringement litigation. (8) There was no difference in the ages of Internet patents and NIPs when they became the subject of litigation — both kinds were about 4.5 years old. (9) Once patent infringement litigation was initiated, the owners of litigated Internet patents were significantly more likely to settle before judgment than the owners of litigated NIPs (especially when probable settlements were taken into account along with obvious settlements, which we believe is the more accurate metric). (10) Across both sets of patents, the larger the number of potential infringers involved in a case (defendants in infringement actions and plaintiffs in declaratory judgment actions), the less likely the case was to settle. (11) Internet patents and NIPs went to trial at about the same rate. (12) When failing to settle, the owners of NIPs won on the merits at a significantly higher rate than did owners of Internet patents — although the win rate for NIP owners was quite low at around 16%, the win rate of Internet patents was even lower by a substantial margin. This finding did not hold up in regression analysis, however; when the effects of other variables were taken into account in a logistic regression analysis, there was no significant difference in the win rate for accused infringers between Internet patents and NIPs. Accused infringers did win more often when Internet patents were asserted against them than win they defended against NIP complaints, but the relatively small number of observations prevented the difference from being statistically significant. (13) Surprisingly, owners of both kinds of patents were significantly more likely to win as the number of inventors on the patents increased. (14) The longer that applications for Internet patents and NIPs had spent in the PTO before issuance, the less likely accused infringers were to win. (15) Accused infringers were less likely to win on the merits when the Internet patents or NIPs asserted against them had been litigated previously. (16) Across both sets of patents, the larger the number of potential infringers involved in a case, the more likely these potential infringers were to win a judgment on the merits. That is, the more infringement defendants per case, they more likely these defendants were to win. (17) There was no difference between the different types of patents in the percentage of cases that were terminated for procedural reasons. We also discussed several other findings of interest.

Patents and Business Models for Software Firms

We analyze the relation between patents and the different business models available to firms in the software industry. The paper builds on Cusumanos work defining the differences among firms that sell products, those that provide services, and the hybrid firms that fall between those polar categories. Combining data from five years of Software Magazines Software 500 with data about the patenting practices of those software firms, we analyze the relation between the share of revenues derived from product sales and the firms patenting practices. Accounting for size, R&D intensity, and sector-specific effects, the paper finds a robust positive correlation between product-based business models and patenting rates. We also present in this draft preliminary results suggesting that there is no significant relation between patenting practices and the extent to which the firms revenues are derived from software products and services, as opposed to hardware or other lines of business.

Self‐policing and reinforcement as alternatives to government regulation for quality assurance in research

The issue of government regulation of scientific research exists in the context of an increasingly litigious society and a growing reliance on legal contracts in lieu of informal agreements. These trends suggest regulation of research may intensify in the absence of countermeasures by the researchers themselves. This paper explores the possibility that increased or altered discipline within the research community itself could result in sufficient quality control to obviate the need for increased government regulation. Possible actions are suggested for program managers, for universities, and for researchers.

Comprehensive audits for use in evaluating scientific research

Calls for new arrangements to deal with problems involving the integrity of results reported from scientific research have generally taken the form of investigations into data quality and the research methods utilized. Alternatives to proposed government regulation of scientific research have included suggestions for using accounting type audits of the kind used for attesting to the representations of management in corporate financial reports. Here a different type of audit is suggested for the advantages it offers in dealing with the situations in science—which are much less structured and not restricted to situations like those to which financial accounting audits are directed. A broadening from “accounting”; to “accountability”; is thereby achieved. Experimentation with the use of such audits is suggested which can simultaneously help to improve scientific processes and accommodate the interests of the public in responding to the results of scientific research. A beginning is also suggested in the form...