For the half-year to 30 June 2014, the IPKat's regular team is supplemented by contributions from guest bloggers Alberto Bellan, Darren Meale and Nadia Zegze.

Two of our regular Kats are currently on blogging sabbaticals. They are David Brophy and Catherine Lee.

Friday, 28 June 2013

More on US patent litigation statistics - correlation with granted patents

Once a Kat always a Kat.  This moggy was delighted to receive an observation from Emeritus Kat Mark Schweizer (now a Rechtsanwalt in the Swiss firm of Meyerlustenberger Lachenal ) in response to the post a few days ago about US patent litigation statistics.

Says Mark:

Regarding your post - I was always wondering how the increase in litigation compared to the increase in patents granted. Turns out, the study answers this, with a quite striking graph on page 6. There is an almost perfect correlation between patents granted and patents litigated. 
Which seems to indicate, to this naive mind, that not much has changed with regards to the decision whether to litigate a patent or not. A small, steady, fraction of granted patents are litigated. That seems to indicate that neither patent quality nor incentives to sue have changed dramatically between 1991 and 2012. It is simply that more granted patents lead to more litigation, which is not exactly surprising. And which is not the public perception, with the brouhaha over NPEs and such.
The graph is this one, which even this untutored moggy can see shows remarkable correlation.



The IPKat of course fully agrees with himself.  How could he not?  But Merpel recalls from her school applied mathematics lessons that in the realm of statistics correlation does not imply causation.  So, is it the existence of the patents that drives the litigation, or does the desire to conduct patent litigation drive the obtaining of patents?  Or are the numbers of patents and amount of patent litigation driven separately by other considerations, which happen to correlate well?  What do readers think is the correct explanation for this extraordinary correlation?

8 comments:

Anonymous said...

It would be interesting to see the same graph for other jurisdictions.....Jackie

Anonymous said...

Surely given that this is *grant* vs litigation, rather than applications vs litigation, there is some lead time between the decision to file more applications and the decision to litigate more (which leads to both moving in almost lock step of the graph). This would seem to suggest that the higher grant rate causes higher litigation, rather than vice versa.

However, a quick look at the scales shows that grant goes up about 2.5x, over the period and litigation goes up about 5x, so litigation is still rising over twice as fast a grant (and the graph is somewhat misleading).

An alternative (though unlikely) explaination would be that weaker patents are being granted (so the growth in applications is less than that of granted patents), which causes defendents to contest validity more often rather than settling. Though that's a good examply of how you can prove anything with statistics (e.g. storks bring babies, since houses with more children in regions with storks are more likely to have storks nesting).

Peter Smith said...

To me, the two scales are highly misleading. They are not in proportion to one another (their zeroes would not be at the same level) and they appear to have been chosen so that both curves fit neatly into the available space. Effectively the curves have been forced to coincide at both ends so it is only the correlation of the shapes of the curves between the end points that is relevant.

It appears that the ratio of cases filed to patents granted was approximately 1.9% in 2012 and 1.2% in 1991. That seems a signifcant change.

To the extent that there is a correlation (e.g. the dip in 2005 and the increase from 2009) it seems strange that there is no time lag, given that not all patent cases are started immediately upon grant of the patent.

Anonymous said...

What is intriguing is that the level of infringement is rising with the rise in patent grant. I am of course assuming that there are infringing activities. :-)

Clearly the threat of expensive patent litigation is not a deterrent.

How about number of assignees v number of infringers?

Anonymous said...

There is an artificial bump induced by the anti-joinder provisions of the AIA that needs to be accounted for.

Anonymous said...

Taking into account the AIA artificial induced bump, and also taking into account that patents do not disappear after one year (and that suits are not immediately filed), a better chart would show the number of suits per year as a percentage of available (active) patents that a suit could be filed against.

I would daresay that such a graph would be anti-climatic to the hoopla and froth that "NPE's are out of control" precisely because it would show a consistent DECLINE (taking into account the anti-joinder anomaly) in the number of cases brought against in comparison to the number of patents available against which such suits could be brought.

Anonymous said...

For instance, assume a constant slope diagonal line starting in the lower left of the graph rising to the upper right of the graph.

Given that AT THE LEAST (for US cases, and I am assuming that the graph is for US cases), 100% of all patents within the first renewal period would still be active, the actual RATE of litigation (patent cased filed divided by patents granted and alive) would DROP from 1% in 1991 to 0.55% in 2012 - and that is NOT EVEN taking into account the false bubble of filings mandated by the AIA (properly removed by the straight line example).

Unfortunately, this has the opposite effect of the desired "Trolls are out of control" message.

As they say, truth hurts. But one must be careful about whose 'truth' is being hurt.

Darren Smyth said...

The Register has, I noticed, just posted on this as well:

http://www.theregister.co.uk/2013/07/08/pwc_2013_patent_suit_study/

Subscribe to the IPKat's posts by email here

Just pop your email address into the box and click 'Subscribe':