News
Intellectual Property
Jul. 31, 2002
Counsel Can Use Statistical Methods to Assess Patent Risk
Focus Column - By James E. Hopenfeld - It has become conventional wisdom - and rightly so - that a strong patent portfolio is a key to corporate success. Patents are, after all, limited monopolies on the technologies that they cover. Patents protect markets, justify research-and-development expenses and can be a prerequisite to obtaining the capital that a business needs to survive.
Focus Column
By James E. Hopenfeld
It has become conventional wisdom - and rightly so - that a strong patent portfolio is a key to corporate success. Patents are, after all, limited monopolies on the technologies that they cover. Patents protect markets, justify research-and-development expenses and can be a prerequisite to obtaining the capital that a business needs to survive. Indeed, patents have become so valuable that some businesses now are devoted solely to developing technology, protecting it with patents and then licensing the patented technology for a fee.
Substantial risks are associated with every patent, however. Should it prove necessary to test a patent in litigation, it may not survive the court's scrutiny. The court could find the patent to be invalid or unenforceable and, therefore, worthless. Even if the court holds the patent valid and enforceable, it could find that the technology that the patent covers is so limited that the corresponding patent monopoly is of little value. Litigation itself is a risk. Win or lose, patent litigation can disrupt business operations and cost millions of dollars.
As patents have become more important, so has assessing these "patent risks." But that has never been an easy task. To the outside observer, the technology is often incomprehensible. The patent laws appear to be a mysterious fog. The resulting risk analysis is maddeningly complex and resists easy answers.
Some patent practitioners have, nonetheless, renewed efforts to demystify patent-risk assessment. To do so, they increasingly have employed a long-neglected tool: statistics. Attorneys now use statistical methods routinely to examine questions such as how often trial courts declare patents invalid or unenforceable, how often they find that a patent covers an accused technology (i.e., infringes the patent's claims), and how often trial court decisions are upheld on appeal.
There has been, in short, a boom in patent-related - indeed, law-related - statistical analysis. In the past few years, statistically based studies have been published in many of the major patent law-oriented journals.
The U.S. Circuit Court of Appeals for the Federal Circuit, responsible for handling appeals in patent cases from District Courts across the entire United States, keeps its own statistics and has from time to time published them at judicial conferences. Data collected by the U.S. government can be accessed on the Internet at http://teddy.law.cornell.edu. No doubt there are further, more extensive studies, in gestation.
Why the sudden interest in statistics? Actually, the interest is not so sudden at all. The reason few statistical studies were published until recently has had more to do with a lack of sufficient data than disinterest.
The only practical way to generate useful patent-risk data is to derive it from the outcomes of trials and subsequent appeals in federal courts. Before the creation of the Federal Circuit in 1982, however, relatively few patent cases went to trial, and jury trials were uncommon. Since the creation of the Federal Circuit, however, the number of patent cases going to trial has steadily increased and jury trials have become commonplace.
In the meantime, the U.S. Patent Office has issued many more patents. In just the past 20 years, it has issued more than 2 million. It had taken more than 100 years, however, to issue the first 2 million. In 2001 alone, the Patent Office granted more than 166,000 patents. In 1981, on the cusp of the information-technology boom, the Patent Office had granted almost 60,000 fewer than in 2001.
With more patents, more patent disputes and more trials, it finally has become possible to do meaningful statistical analyses. In 1998, John Allison and Mark Lemley published what was then one of the most comprehensive studies ever made. Their study examined 300 cases over an eight-year period. John R. Allison & Mark Lemley, "Empirical Evidence on the Validity of Litigated Patents," 26 AIPLA Q. L.J. 185 (1998).
Just four years later, Kimberly Moore published a study examining more than 1,200 cases over a 17-year period. Kimberley A. Moore, "Judges, Juries, and Patent Cases - An Empirical Peek Inside the Black Box," 11 Fed. Cir. B.J. (No. 2) 209 (2002).
Some consistent patterns have emerged from the data. Moore's statistics show, for example, that courts are more likely to find in favor of the patentee (58 percent) and against an accused infringer (42 percent). Juries are more likely to find in favor of patentees than judges are (68 percent to 51 percent, respectively). Judges' and juries' decisions are affirmed at about the same rate (78 percent). The Allison and Lemley data shows similar patterns.
You might think that you could rely on these statistics to evaluate patent risk. After all, the data is based on sufficiently large statistical samples and appear relatively consistent from study to study.
Yet these broad, rough cuts with the numbers can be misleading. Add just one additional variable, and the data appears to tell a very different story. Allison and Lemley considered, for example, the effect that the type of technology had on the likelihood that a patent is declared invalid. Over all, they had found approximately 54 percent of patents were found by courts to be valid (note that this is very close to Moore's 58 percent). But they also found that pharmaceutical (73 percent) and computer-related patents (65 percent) were somewhat more likely to be found valid, while electrical (53 percent) and software (33 percent) related patents were somewhat less likely to be found valid.
Allison and Lemley found even greater disparities when they considered where the patent case was tried. Viewing data from 25 of more than more than 100 U.S. District Courts, they found that patent validity success rates exceeded 80 percent in five of them. Indeed, in the eight cases from the Western District of Wisconsin, all of the patents were found to be valid.
There were, on the other hand, four districts in which patents were found to be valid no more than 20 percent of the time. In two, the District of New Jersey and the Western District of North Carolina, none of the patents (out of 13) was found to be valid.
Taking into account additional variables should, in theory, make statistically based patent-risk assessment somewhat more reliable. But that, too, is deceptive. Once variables such as technology type and geography are factored in, the relevant data sets from which statistics are drawn become significantly smaller.
Allison and Lemley's statistics showing that software patents have been found to be valid only 33 percent of the time are based, for example, on a sample size of just three reported cases. Had just one case turned out differently, the reported value would have been 33 percent higher or lower. If variables such as technology type and geography are to be considered - or indeed any variable that drastically cuts down on sample size - much more data, and perhaps many more years of data collection, is needed.
Putting aside the problem of ignoring seemingly important variables like technology type and geography, the data itself may be fundamentally flawed. The studies are, by necessity, limited to data generated from patent cases that were litigated. But most patents never taste litigation, and of those that do, in most cases, the litigation settles before the court finally resolves the validity or infringement of the patent.
Despite some flaws, statistics can be useful. They can be used to provide baselines for more sophisticated analyses of patent risk or for a rule-of-thumb analysis where a more detailed evaluation doesn't make sense. And they have uses unrelated to the assessment of patent risk at all, such as assisting docket management in federal courts.
It is highly unlikely that statistical analysis will replace solid legal evaluation of patent risk. It should, however, become an increasingly useful part of the lawyer's toolkit. To make the best use of this new tool, however, lawyers should consider becoming more familiar with its strengths and weaknesses.
James E. Hopenfeld is an attorney at the Palo Alto office of Fish & Neave. His practice focuses on intellectual property litigation and counseling.
By James E. Hopenfeld
It has become conventional wisdom - and rightly so - that a strong patent portfolio is a key to corporate success. Patents are, after all, limited monopolies on the technologies that they cover. Patents protect markets, justify research-and-development expenses and can be a prerequisite to obtaining the capital that a business needs to survive. Indeed, patents have become so valuable that some businesses now are devoted solely to developing technology, protecting it with patents and then licensing the patented technology for a fee.
Substantial risks are associated with every patent, however. Should it prove necessary to test a patent in litigation, it may not survive the court's scrutiny. The court could find the patent to be invalid or unenforceable and, therefore, worthless. Even if the court holds the patent valid and enforceable, it could find that the technology that the patent covers is so limited that the corresponding patent monopoly is of little value. Litigation itself is a risk. Win or lose, patent litigation can disrupt business operations and cost millions of dollars.
As patents have become more important, so has assessing these "patent risks." But that has never been an easy task. To the outside observer, the technology is often incomprehensible. The patent laws appear to be a mysterious fog. The resulting risk analysis is maddeningly complex and resists easy answers.
Some patent practitioners have, nonetheless, renewed efforts to demystify patent-risk assessment. To do so, they increasingly have employed a long-neglected tool: statistics. Attorneys now use statistical methods routinely to examine questions such as how often trial courts declare patents invalid or unenforceable, how often they find that a patent covers an accused technology (i.e., infringes the patent's claims), and how often trial court decisions are upheld on appeal.
There has been, in short, a boom in patent-related - indeed, law-related - statistical analysis. In the past few years, statistically based studies have been published in many of the major patent law-oriented journals.
The U.S. Circuit Court of Appeals for the Federal Circuit, responsible for handling appeals in patent cases from District Courts across the entire United States, keeps its own statistics and has from time to time published them at judicial conferences. Data collected by the U.S. government can be accessed on the Internet at http://teddy.law.cornell.edu. No doubt there are further, more extensive studies, in gestation.
Why the sudden interest in statistics? Actually, the interest is not so sudden at all. The reason few statistical studies were published until recently has had more to do with a lack of sufficient data than disinterest.
The only practical way to generate useful patent-risk data is to derive it from the outcomes of trials and subsequent appeals in federal courts. Before the creation of the Federal Circuit in 1982, however, relatively few patent cases went to trial, and jury trials were uncommon. Since the creation of the Federal Circuit, however, the number of patent cases going to trial has steadily increased and jury trials have become commonplace.
In the meantime, the U.S. Patent Office has issued many more patents. In just the past 20 years, it has issued more than 2 million. It had taken more than 100 years, however, to issue the first 2 million. In 2001 alone, the Patent Office granted more than 166,000 patents. In 1981, on the cusp of the information-technology boom, the Patent Office had granted almost 60,000 fewer than in 2001.
With more patents, more patent disputes and more trials, it finally has become possible to do meaningful statistical analyses. In 1998, John Allison and Mark Lemley published what was then one of the most comprehensive studies ever made. Their study examined 300 cases over an eight-year period. John R. Allison & Mark Lemley, "Empirical Evidence on the Validity of Litigated Patents," 26 AIPLA Q. L.J. 185 (1998).
Just four years later, Kimberly Moore published a study examining more than 1,200 cases over a 17-year period. Kimberley A. Moore, "Judges, Juries, and Patent Cases - An Empirical Peek Inside the Black Box," 11 Fed. Cir. B.J. (No. 2) 209 (2002).
Some consistent patterns have emerged from the data. Moore's statistics show, for example, that courts are more likely to find in favor of the patentee (58 percent) and against an accused infringer (42 percent). Juries are more likely to find in favor of patentees than judges are (68 percent to 51 percent, respectively). Judges' and juries' decisions are affirmed at about the same rate (78 percent). The Allison and Lemley data shows similar patterns.
You might think that you could rely on these statistics to evaluate patent risk. After all, the data is based on sufficiently large statistical samples and appear relatively consistent from study to study.
Yet these broad, rough cuts with the numbers can be misleading. Add just one additional variable, and the data appears to tell a very different story. Allison and Lemley considered, for example, the effect that the type of technology had on the likelihood that a patent is declared invalid. Over all, they had found approximately 54 percent of patents were found by courts to be valid (note that this is very close to Moore's 58 percent). But they also found that pharmaceutical (73 percent) and computer-related patents (65 percent) were somewhat more likely to be found valid, while electrical (53 percent) and software (33 percent) related patents were somewhat less likely to be found valid.
Allison and Lemley found even greater disparities when they considered where the patent case was tried. Viewing data from 25 of more than more than 100 U.S. District Courts, they found that patent validity success rates exceeded 80 percent in five of them. Indeed, in the eight cases from the Western District of Wisconsin, all of the patents were found to be valid.
There were, on the other hand, four districts in which patents were found to be valid no more than 20 percent of the time. In two, the District of New Jersey and the Western District of North Carolina, none of the patents (out of 13) was found to be valid.
Taking into account additional variables should, in theory, make statistically based patent-risk assessment somewhat more reliable. But that, too, is deceptive. Once variables such as technology type and geography are factored in, the relevant data sets from which statistics are drawn become significantly smaller.
Allison and Lemley's statistics showing that software patents have been found to be valid only 33 percent of the time are based, for example, on a sample size of just three reported cases. Had just one case turned out differently, the reported value would have been 33 percent higher or lower. If variables such as technology type and geography are to be considered - or indeed any variable that drastically cuts down on sample size - much more data, and perhaps many more years of data collection, is needed.
Putting aside the problem of ignoring seemingly important variables like technology type and geography, the data itself may be fundamentally flawed. The studies are, by necessity, limited to data generated from patent cases that were litigated. But most patents never taste litigation, and of those that do, in most cases, the litigation settles before the court finally resolves the validity or infringement of the patent.
Despite some flaws, statistics can be useful. They can be used to provide baselines for more sophisticated analyses of patent risk or for a rule-of-thumb analysis where a more detailed evaluation doesn't make sense. And they have uses unrelated to the assessment of patent risk at all, such as assisting docket management in federal courts.
It is highly unlikely that statistical analysis will replace solid legal evaluation of patent risk. It should, however, become an increasingly useful part of the lawyer's toolkit. To make the best use of this new tool, however, lawyers should consider becoming more familiar with its strengths and weaknesses.
James E. Hopenfeld is an attorney at the Palo Alto office of Fish & Neave. His practice focuses on intellectual property litigation and counseling.
#337203
Columnist
Daily Journal Staff Writer
For reprint rights or to order a copy of your photo:
Email
Jeremy_Ellis@dailyjournal.com
for prices.
Direct dial: 213-229-5424
Send a letter to the editor:
Email: letters@dailyjournal.com