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.

Monday, 18 March 2013

Patent filing and public debt: here's introducing Schweizer's Law


Whatever happened to Mark Schweizer?  One of the funniest, spikiest Kats, Mark retired from active Kat-service a few years ago when the demands of the real world kept taking him away from this corner of the blogosphere.  But Mark is not forgotten -- and nor has he forgotten us.  Out of the blue, this Kat was overjoyed to receive the diagram depicted above, together with the following text:
"I've had some fun with the EPO statistics 2012 and produced the attached graph.  The graph shows the relationship between public debt as % of GDP (2011 data from Wikipedia) on the x-axis, and number of patent applications per capita for each EU Member State (according to country of residency of first named applicant, according to the EPO's Statistics and Trends here) on the y-axis. 
I've multiplied the per capita filings by 1,000 because the numbers are very small otherwise. I've also excluded all EU Member States with fewer than 100 patent applications filed in 2012 because, with such low numbers, the figures become unreliable (in case you miss Greece -- it had just 79 applications in 2012, so did not make the cut). I've included, on the other hand, the two non-EU Member States USA and Switzerland (CH) for comparison purposes. The correlation of public debt as % of GDP and applications per capita is with R squared = 0.3 is moderate, but not too bad for social sciences. Of course, correlation is not causation, and that is certainly true here. Anyone may interpret the graph as he or she sees fit - I just find it, well, interesting. And fun".
The IPKat, who still pines a little bit for the comforting thoughts of Katonomist Nicola Searle, was greatly cheered by these figures.  Merpel is still pretending to understand them, even though the IPKat caught her reading them upside down.  But what do readers think? Has Mark discovered an important truth? Should we call it Schweizer's Law?

10 comments:

Anonymous said...

Correlation is indeed not causation, but even in case there would be a causation, one could ever wonder which is the cause and which is the consequence...
This triggers interesting considerations when you think about it, but I honestly can hardly draw any conclusion out of it !

Anonymous said...

1) Correlation is not causation.

2) That's a pretty poor best fit line. Cut out LU and CH as anomalous outliers, and there's practically no correlation.

3) On the whole, a very nice example of why one should be cautious of regression analysis.

(oblig. XKCD: http://xkcd.com/552/)

Anonymous said...

The order of the main sequence from LU to IT is the order of the nation's propensity to have things done properly at one end to disorder and just getting on with it at the other. Filing a patent and keeping your budget under control are both orderly things to do. Departures from that it tell us what we know already which is that DE files too many patents and UK not enough. (But what are CH on?)

The sequence from CZ to HU shows that they just do not do patents at all - so retry the correlation coefficient without them. Since PL and ES don't file for patents it is no surprise that they don't want to play in the EU patent.

Anonymous said...

If it is Schweizer's law why is der Schweiz an outlier??

Nicola said...

I completely agree with Mark that Schweizer's law is interesting and fun.

This could be a spurious correlation (i.e. something that suggests a relationship but is actually just coincidence.) Or it could suggest a relationship. Do patent filings reduce national debt? Are countries with high levels of debt also experiencing low levels of patentable innovation?

It's a relatively small sample size, so like an economist, I'd call for more research!

Nathan Wajsman said...

As another economist, I am completely with Nicola here regarding the need for further research. But on first look, this does seem to be a spurious correlation, and also a bit of an artifact of country selection--what would happen if Japan, with its 200% debt/GDP ratio, had been included? When I was teaching in the late 80s, I used to use two examples of "interesting" correlations: (1) the positive relationship between the number of nesting storks and the birth rate in Denmark; and (2) the equally positive correlation between teacher salaries and liquor sales in some US states.

Fede said...

as mentioned, there are some outliers. furthermore, it needs to include time: a cross section analysis is not sufficient, in my opinion. Many countries might have quite a constant behaviour independent from time changes in public debt-gdp ratio.
And finally, literature has found correlation between number of filings and gdp.
however, an interesting new perspective

Anonymous said...

I don't think the theory travels well. Most countries in the developing world have far lower public debt, and yet don't feature in any patenting stakes. Even insofar as Europe is concerned, I suspect that setting a minimum patenting threshold for consideration conveniently allows a huge number of countries to be ignored. Did the below-threshold countries tend to have higher public debt?

Meldrew said...

Patent filings are a luxury good and correlate very well with GDP/head. There should be a very good correlation with sales of designer shoes, handbags, and champagne.

Wait a minute - do sales of designer shoes, handbags, and champagne correlate well with public debt?

I'm confused.

Anonymous said...

This is a dangerous curve - take away the two points on top left, and the regressions would be unlikely to be statistically significant.

Come on guys, we are supposed to be smart to be in the world of I.

Putting the doubtful stats aside..dunno. Would need to see more data.

Subscribe to the IPKat's posts by email here

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