count patents for german patent office

[Alex_G]

Dear all,

we are using PATSTAT (Version Oct2013) in a MS SQL Server environment and trying to reproduce the yearly numbers of patent applications for German Patent Office (DPMA) we found at WIPO IP Statistics Data Center http://ipstats.wipo.int/ipstatv2/index.htm?tab=patent, to become familiar with PATSTAT.

You can see these official numbers for total patent applications (direct and PCT national phase entries) at German Patent Office in the table below in column “pats_dpma”. In column “pats_patstat” you can compare the numbers of our PATSTAT-Query, which seem to be quite different, especially for the years before 2004. We are using following SQL-Code for reproducing the patent counts:

Code: Select all

SELECT DISTINCT 
appln_auth, 
year(appln_filing_date) as year,  
count(*) over (partition by year(appln_filing_date)) as patents

FROM patstatoct2013.dbo.tls201_appln

WHERE appln_auth = 'DE'
    AND appln_kind in ('A', 'W')
    AND year(appln_filing_date) between 1990 and 2010

year | pats_dpma | pats_patstat | diff
1990 | 39329 | 77190 | 196,3%
1991 | 40040 | 76181 | 190,3%
1992 | 41323 | 78161 | 189,1%
1993 | 41747 | 79772 | 191,1%
1994 | 43976 | 82547 | 187,7%
1995 | 46158 | 85665 | 185,6%
1996 | 51833 | 93294 | 180,0%
1997 | 55729 | 99950 | 179,4%
1998 | 57366 | 105845 | 184,5%
1999 | 59531 | 110703 | 186,0%
2000 | 62142 | 116047 | 186,7%
2001 | 60475 | 114098 | 188,7%
2002 | 58187 | 106924 | 183,8%
2003 | 58481 | 102429 | 175,1%
2004 | 59234 | 57574 | 97,2%
2005 | 60222 | 56367 | 93,6%
2006 | 60585 | 54324 | 89,7%
2007 | 60992 | 54568 | 89,5%
2008 | 62417 | 55639 | 89,1%
2009 | 59583 | 52323 | 87,8%
2010 | 59245 | 52691 | 88,9%


We also tried to count patent applications over priority patents or patent families, but there is always a big difference between official numbers and our query-results.

Does anyone have an idea, why these numbers are so different? Is there a mistake in our SQL-Query, are we using a wrong “counting-method” or is there another explanation?

We are grateful for any hint.

Many thanks,
Alex

[mkracker]

Like you already did, it is good practice to check the quality of your data and the plausibility of your result. Basically your query is fine, but you did not consider some particularity of the German patent office. The DPMA also publishes granted EPs in their Bulletin using PUBLN_KIND code D1. These additional publications - or their related EP applications to be more precise - distort your numbers. Obviously, this publication practice at the DPMA changed in 2004 - but I did not follow up on that.

So the query I would use is:

Code: Select all

select year(appln_filing_date), count(distinct a.appln_id)
from tls201_appln a
join tls211_pat_publn p on a.appln_id = p.appln_id
where appln_auth = 'DE'
and appln_kind =  'A'        -- national applications and PCT applications in the national phase
and publn_kind <> 'D1'     -- exclude applications with publications for "Granted EP published in the Bulletin"
and year(appln_filing_date) between 1990 and 2010
group by year(appln_filing_date)
order by year(appln_filing_date)

I excluded the APPLN_KIND = 'W', because 'W' would include the international phase of the patent. So in fact I exclude all applications where the DE office was the Receiving Office of a PCT.
I excluded PUBLN_KIND = 'D1', as discussed above. To be on the safe side, also check the remaining publication kind codes, whether you want them included or not. A list of all publications kind codes and their meaning can be found at the EPO site at the "Useful tables and statistics" section in http://www.epo.org/searching/data/data/ ... gular.html.

As a result I get over the years consistently about 78% to 85% of the number you provided as reference. I strongly assume these 15-20% applications "missing" in PATSTAT are applications which never have been published. The DE office is counting them, but without publication these applications never make it to the EPO, and consequently to PATSTAT.