If we bother them more, are they less cooperative?

Download Report

Transcript If we bother them more, are they less cooperative?

DATA UTILITY, CONFIDENTIALITY, AND THE
PRODUCTION-POSSIBILITY FRONTIER:
STRIKING A DELICATE BALANCE
Daniel Beckler
United States Department of Agriculture
National Agricultural Statistics Service
Timothy Mulcahy
NORC at the University of Chicago
Topic (ix): Statistical disclosure limitation for table and analysis servers: how to make outputs of
modern data access infrastructures safe
UNECE/Eurostat Work Session on Statistical Data Confidentiality
Tarragona, Spain • 26-28 October 2011
Slide
1 1
Slide
Overview of Microdata Dissemination Techniques
 Public Use Files
 Online Statistical Data Cubes and Tabulation
Engines
 Remote Batch Processing
 Synthetic Microdata
 Remote and Physical Data Enclaves
 With these methods, there is a trade-off between
disclosure risk, the amount of analytic utility, and
the ease of access.
UNECE/Eurostat Work Session on Statistical Data Confidentiality
Tarragona, Spain • 26-28 October 2011
Slide
1 2
Slide
National Agricultural Statistics Service
 United States Department of Agriculture
 Conducts censuses & surveys on U.S.’s farm
population.
 Generates official USDA agricultural statistics, many
impact global commodity markets
 Paper discusses how NASS protects the
confidentiality of microdata, while providing as
much analytical utility as possible to the users of
the official statistics as well as researchers.
UNECE/Eurostat Work Session on Statistical Data Confidentiality
Tarragona, Spain • 26-28 October 2011
Slide
1 3
Slide
United States Census of Agriculture
 Conducted every 5 years. Produces very detailed
data at the U.S., state, and county (i.e., sub-state)
levels.
 Data for individual agricultural operations are
protected from disclosure in published totals by
using a threshold rule and a dominance rule
 Primary suppressions result directly from these
rules
 Complementary suppressions are then determined
to ensure primary suppressions may not be
calculated from published data.
UNECE/Eurostat Work Session on Statistical Data Confidentiality
Tarragona, Spain • 26-28 October 2011
Slide
1 4
Slide
United States Census of Agriculture
 Loss of utility of the Census due to suppressions:
Domain
Overall Count of
Estimates
Number of
Primary
Suppressions
Number of
Complementary
Suppressions
Total
Total Number of
Suppressions as
Suppressions
% of Estimates
US
29,075
255
351
606
2.08
State –Low %
61,000
8,614
2,721
11,335
18.58
State – High %
16,095
5,111
2,195
7,306
45.39
2,556,586
430,843
151,506
582,349
22.78
All (US & State)
UNECE/Eurostat Work Session on Statistical Data Confidentiality
Tarragona, Spain • 26-28 October 2011
Slide
1 5
Slide