Page 328 - Invited Paper Session (IPS) - Volume 1

IPS155 Laura B.
external researchers, while preserving confidentiality, where needed (Bruno,
D’Aurizio and Tartaglia-Polcini, 2011). Households’ survey data are under the
general protection of the GDPR, while GDPR rules do not apply to companies,
therefore no limit exists, in theory, for the dissemination of firms’ elementary
data. Contrarily, Banca d’Italia has chosen to not disseminate firms’ granular
data, because our business surveys include information that companies may
not want to disclosure, as expected turnover, investment plans, etc.; therefore,
in order to preserve a high response rate, we have chosen to not disseminate
firms’ elementary data. Indeed, Banca d’Italia makes it clear to her surveys’
respondents that information provided through business surveys will not be
diffused to the public in an elementary format. The promise of this additional
safeguard is crucial for the collection of precious information that would not
be elsewhere available.

3. Firms’ survey data
Having voluntary chosen not to disseminate firms’ elementary data, we
had to find a way to safeguard confidentiality. Different options were available.
Anonymization is the first, easy possibility. Still we have chosen not to go this
way as we consider it an insufficient protection for firms’ data, as researchers
could always trace back a respondent firm by simultaneously using
stratification dummies (region, size, sector). It could also then be possible to
expunge these variables from the database, but by exactly matching a
particular data (as for example turnover), re-identification of the respondent
could in theory still be possible. Another possibility was the use of
confounding techniques, but, again, we valued these techniques as not
completely safe, because the algorithm used for confounding data could be
identified. We neither chose the option usually adopted for data with the
highest confidentiality restrictions, i.e. providing access to elementary data in
a devoted laboratory (also called “data enclave”), where computers don’t allow
the user to take away any data if not in an aggregated form, and laboratory
employees check the result of the elaborations and their non-violation of
confidentiality restrictions. A laboratory significantly reduces utility, because
the researcher has to go in person to the devoted location. On the contrary,
willing to disseminate granular data to external economists and having
adequate technical devices, we chose remote processing. Remote processing
allows researchers to process elementary data, ensuring that individual
information cannot be visualized.

317 | I S I W S C 2 0 1 9