Page 320 - Special Topic Session (STS) - Volume 1

STS441 Andrea N.
Similar results were found by Neri and Monteduro (2013) who carried out
an adjustment of housing wealth based on the aggregate distributions of
ownership from tax records. The authors found that SHIW underestimates
both the number of taxpayers who own just one and those who own more
than five units of housing. Correcting the SHIW data by aligning the sample
data with the administrative data increases total housing wealth by about a
quarter.
As far as financial wealth is concerned, Cannari, D'Alessio, Raimondi and
Rinaldi (1990) and Cannari and D'Alessio (1993) performed a statistical
matching of the financial assets declared by SHIW respondents with data
provided by a sample of commercial bank clients from a survey carried out by
the bank. The authors used statistical matching to model non-reporting and
under-reporting behavior and to adjust SHIW data. A similar approach was
used (D'Aurizio et al., 2006). The adjusted estimates of financial assets average
more than twice the original figures, reaching 85 percent of the aggregate
coming from financial assets. The paper also adjusted financial liabilities,
whose corrected values are on average about 40 percent higher.
Other studies were mainly focused on the analysis of the differences in
definitions and concepts between the micro and macro sources (Antoniewicz
2005 and Bonci et. Al 2005).
More recently, D’Alessio and Neri (2015) conducted several adjustment
experiments on SHIW data, combining different imputation and calibration
techniques, in order to produce estimates consistent with the macro-
economic information available from other sources. The study shows some
results are robust to the adjustment method applied. For instance, whatever
the method is used, the adjusted estimates of the Gini concentration indexes
of both income and wealth are always higher than the unadjusted one.
Yet, the authors also show that results are strongly affected by the auxiliary
information available. Without any external information, they are basically
driven by the choice of the assumptions behind the adjustment models.
The main limitation of all these studies is that they were not based on an
exact matching of survey data with administrative records. This was mainly due
to the existence of legal constrained that made it impossible for data
producers to share their administrative records with Banca d’Italia. Given such
limitation, the only solution available was to use statistical matching
techniques whose goal is to find the most similar observations in two different
databases (based on the observable characteristics). Banca d’Italia is currently
working to change this situation for the future.

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