Access to individual data in France Michel ISNARD Insee – Head of Legal Affairs 28/10/2013

Download Report

Transcript Access to individual data in France Michel ISNARD Insee – Head of Legal Affairs 28/10/2013

Access to individual data in
France
Michel ISNARD
Insee – Head of Legal Affairs
28/10/2013
Individual data files in France
• Public Use Files
• Scientific Use Files
• Secure Use Files
• Specific topics
2
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality – Ottawa 2013
28/10/2013
Public Use Files
• On Insee’s website
• “Households” data
 Labour force survey
 Census data : 2 files
One with a localisation at regional level (27 regions in France) and
detailed social variables
One with a localisation at municipality level and variables with
aggregated modalities
 Some register files
• http://www.insee.fr/fr/bases-de-donnees/fichiers-detail.asp
 In French
3
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013
Scientific Use files
• For researchers with specific documentation for
researchers
• But :
 Who is a researcher ? And who is not ?
 What kind of documentation did they need ?
• Statisticians need some help
4
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013
Scientific Use Files (2)
• Réseau Quetelet : French Data Archives
 Formally created in 2001
 But result of a longer cooperation between Insee and some
researchers
• Disseminates Insee (and other) SUF to French and
foreign researchers.
• Therefor determines who is a researcher or not.
• Help Insee to create a documentation usable by
researchers
• http://www.reseau-quetelet.cnrs.fr/spip/
5
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013
Confidential Files
• Long history in France for business data
 Since 1984
• More recent for Household data
 Since 2008
• Procedure :
 Opinion by an external committee : Statistical Confidential
Committee
Chaired by a judge
Participation of representatives of business unions, worker unions
and researchers
 Agreement of Insee
 Decision by National Archives
6
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013
Confidential files (2)
• Longer procedure than in other countries
 But probably more acceptable
 200 access requests a year
• Access Through Genes’s CASD
 http://www.casd.eu
7
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013
How to get data ?
• First stop : Réseau Quetelet
 If SUF enough, get the data
• Second stop :
 Confidentiality Committee secretary and the data producer
 To see if confidential data will solve the problem
• Third stop
 Confidentiality Committee
• Fourth stop
 CASD
8
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013
Specific topics
• Output checking – My OWN PERSONAL OPINION
 Is it useful ? Enough ? Efficient ?
 Will only cope with remote access
9
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013
Output checking in remote access
• Some preliminary remarks :
 An output file can’t be more informative than the
confidential file the researcher is allowed to browse
 A researcher has already signed a confidentiality clause
and could be, depending on national laws, bound by penal
responsibility
 A researcher could easily remember the value of some
specific variable and therefore extract it from the safe
centre.
 Who is in charge if there’s a confidentiality break ? The NSI
? The researcher?
10
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013
Output checking in remote access
• OC can’t be effective :
 If a researcher wants to smuggle ONE specific information
outside the secure centre, NSI can’t check. He/She just has
to remember it!!!
 He/She could also makes specific operations to know some
confidential data about a group of units.

 Checking thoroughly all the output of a researcher and are
sure there’s no confidentiality breach is not enough
 You also have to check them with every published output
made on the same data
11
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013
Output checking in remote access
• OC could be very expensive :
 Of course, we could have the researchers paying OC, but is
it a long term solution ?
 Specially if 99% of researchers follow strictly confidentiality
rules
• OC is very dangerous for NSIs :
 If an individual person or a business happens to know
about some confidentiality breach, the NSI in charge of
then OC could be accused and confidence could be lost
• But we need to have a protection against a complete
download of the data :
 Look at the size of the output
 Check its form
12
Joint UNECE/Eurostat Work Session on Statistical Data Confidentiality - Ottawa 2013
28/10/2013