navigational aids:
News ticker:
topics covered:
|
12 May 06.
In many of my little writings, I have taken as given that patents are a good thing. They foster innovation. And y'know, I basically believe this for mechanical and chemical inventions.
But I have no empirical evidence. I've searched the economics literature over and over again, and I have not found a single paper that provides empirical evidence for the statement patents foster innovation. There are enough logical arguments and models one way or the other, including my own, and everybody seems to generally have some idea about it, but hard data is 100% lacking.
Much of the problem is the simple issue of measuring innovation. For most studies of the form Does (fill in blank) foster innovation?, the researchers just use patent counts as a proxy for innovation, so they are basically asking Does (fill in blank) foster patents? Typically, they look at patents weighted by citations, in an attempt to ferret out the important patents from the more absurd cases. This too has its flaws, but whatever. But clearly, using patent counts as a proxy for innovation ain't gonna work when patent laws are the independent variable. We could count R & D expenditure, but that's rarely aggregated on a national scale, and company-level definitions and reporting preferences basically make the reporting garbage anyway.
Say that one Thursday morning, designs of word games become patentable. You figure the word game industry is small enough, and you could consistently total up R & D for every company in the word game industry before and after the change in law. All well and good, but what about substitution away from board game and publishing method and billboard building R & D? Your study measuring one small subset of the economy says nothing about whether total R & D spending is going up or down.
And hey, R & D does not equal innovation, as the solar power researchers have (sadly) shown us. Maybe we should be looking at rubber-to-the-road product introductions, but measuring that is even harder. Does the one in pink count as a new product? What about the one that has features disabled so it can sell at a lower price point?
Having established that the lack of definite measurement will guarantee that our data doesn't measure what we want, we can move on to the statistical problems in analyzing that data. In the ideal experiment, you change one variable, leave everything else constant, and measure the effect on the dependent variable. But this is rarely an option for a law. Option A would be to do a cross-country comparison. The regression shows that strong patent laws definitely lead to higher GDP, which is all very convincing until you realize that the only countries with strong patent laws are the USA, the EU members, and Japan, so the regression translates to saying that being the USA, EU, or Japan is good for growth. These regions are certainly doing OK for themselves, but there's no convincing regression that will prove that it is because of patents, because so very much changes from one country to the next.
Option B is the time series approach. The problem here is just like the cross-country problem: from one year to the next, everything in the economy is changing at once, so a jump could be due to a change in law or due to a hundred other factors. Sometimes there's a smoking gun: you graph the number of private schools, and it takes a sharp spike right when desegregation passed into law. [Quick---somebody send me a graph of this!] But there is no smoking gun for patents, because (1) shifts in patent law are never night-and-day (even at the founding of the US Patent Office, which sort of trickled into existence) and (2) shifts in R & D have never shown themselves to be particularly night-and-day anyway.
Oh, and to finish off Option B, we have the long-term question. Perhaps patents initially draw many firms into a field, but eventually, we find that one or two firms manage to score all the patents and then hold an effective monopoly that causes the system to shut down. So we need to look at innovation over the long term, which is again a tough one for our time series analysis tools to tackle.
So those are the hurdles that an empirical paper must surmount before it can claim that patents foster innovation, and I have never seen a paper that came even vaguely close. But I have never made the statement It's not in the literature and actually been right. So I turn to you, my dear readers. From those of you I know, you tend to be academic and smarter than the average Netizen, and if you've read this far you have some interest in IP. So do you have any empirical papers that convinced you of the value of patents? Do you have any innovative suggestions on how to measure innovation?
[link] [A comment]
[Previous entry: "The schism"]
[Next entry: "The Web as human network"]
Replies: A comment
|
on Friday, May 12th, Andy said
What about using productivity measures for innovation? Suppose there is no innovation. Then I can imagine productivity increasing for only two reasons: practice (aka learning by doing) and economies of scale as the country grows. I think that you could probably correct for both of these by including a time trend and the overall size of the industry. Everything else is almost by definition the result of innovation. That would of course include managerial innovations like Taylorism, TQM, etc., although I also think that those are legitimately innovative.
The only other possible problem is the importation of innovation -- American car manufacturers got a lot more productive after the Japanese showed them how to do it. However, the Japanese were genuinely innovative in the first place, so maybe you could use the timeliness of productivity growth to measure who first came up with the idea.
But I think that you should look into the productivity literature, probably someone else has had this idea before, but maybe the patent folks haven't cottoned onto it yet (or have they?).
|
|
