Anyone inclined to read this book should just ignore its subtitle -- "How Judges, Bureaucrats, and Lawyers Put Innovators At Risk." I can only imagine that was thrown in to make this book appeal to right-wingers. Which is really strange, come to think of it: the book is sedate, scholarly, reasoned, and exhaustive, and if anything it will appeal to readers who enjoy Larry Lessig and Richard Posner. In no way do Bessen and Meurer suggest that the system be scrapped and those "Judges, Bureaucrats and Lawyers" be thrown out. In the world Bessen and Meurer desire, "bureaucrats" will still examine patent claims; lawyers will still prepare patent applications; and judges will still hear infringement claims. Bessen and Meurer want to change policies here and there to make the system more efficient, but it will still be the same system by the time they're done.
The dispute over patents has become rancorous in the era of the Internet, with lots of loud people asking that it be torn down, lots of other people claiming that the first group are unwitting pawns of big business, etc. Bessen and Meurer avoid almost all of that. Their claim is straightforward, and avoids all the parts of the debate that are unnecessary for establishing their point -- which is simply this: that as a system of property, there are certain things we should expect of a patent system. That is, just as real property gives you certain rights over, say, a parcel of land, and the right to exclude ne'er-do-wells from that land, so a patent is a legal right to exclude people from using your invention. And there are certain things that we should expect of patents, just as we would expect them from property law. First among these is notice: it should be easy to determine whether you've trespassed on my land, and just as easy to determine whether you've violated my patent.
In a system where notice is working perfectly, we'd expect very few patent-infringement lawsuits. In some fields, say Bessen and Meurer, that's exactly what we see: the chemical and pharmaceutical industries have very low rates of infringement, because comparing your small molecule to a patented small molecule is easy. When the product gets more complex and abstract -- as with a computer algorithm -- deciding whether you've infringed gets more and more difficult, and the number of suits balloons.
Still, if lawsuits to defend against infringement claims were cheap enough, patents would be a net economic positive: the cost of lawyers when you accidentally infringe would be less than the money you bring in from non-infringing patented technologies. By this measure, Bessen and Meurer say that patents stopped being a net economic positive -- outside of the chemical and pharmaceutical industries -- in the mid-nineties.
In support of their conclusions, they use an arsenal of statistical methods based on the stock prices of public companies. Everyone knows the defects in using the stock market to infer knowledge, and Bessen and Meurer are not fools. Still, the most reliable data they could find were stock-price data, on which they conducted event studies to determine what value the market attached to particular patents and infringement lawsuits. Absent any data from lawyers or companies about how much they're charging or paying, this is probably the best one could hope for.
Their second method of estimating value is to study rates of patent renewal. 20% of patents expire after four years; 21% expire after eight; 17% expire after 12; and 42% last the full 17-year term. Based on the cost of renewing, we can estimate how much the owner values the patent. Seeing that 58% of patents expire by their 12th year, and that $2,327 in renewal fees will have been due by then, we can estimate that the median patent is worth less than $2,500. Similar, but obviously more sophisticated, analyses are spread throughout Patent Failure; they constitute its bulk.
Bessen and Meurer blame the decreasing value of patents on a number of different factors. One is the rise of software and business-method patents, which are inherently abstract. They give a fascinating account of the Karmarkar patent, which covers a certain interior-point method for solving linear-programming problems. Years after Karmarkar obtained his patent, mathematicians discovered that the Karmarkar algorithm was formally identical to another that had been been in wide use for years. This sort of equivalence is much more likely for an abstract patent such as one governing a mathematical algorithm, and makes knowing ahead of time whether you've infringed that much more difficult. Abstraction harms the notice function of patents, hence harms their functioning as property.
They sidestep any number of other questions, such as the cost of "patenting around" -- that is, developing new technologies only because existing technologies are all patented. Adding in these costs would likely only help their case, though one would have to crank through the numbers to be sure.
My only concern about this book is its subtitle. If I were to issue a second edition, I would drop the populist-sounding subtitle and rename it to what it really is: "A Calm Look At The Evidence."
