Page 431 - Special Topic Session (STS) - Volume 3

STS552 João Falcão Silva et al.
at aggregate level in the presence of confidentiality restrictions. The
consequence is that aggregate estimations are broad based, less precise and
may not provide good results for distant historical quarters, neither to
estimate bilateral positions. The gap in amounts between estimated figure
from ‘all reporting countries’ and that from sum of leaf-level reported plus
estimated amounts is narrowed down when the coverage for a counterparty
country is high as in the latest quarters for Portugal (coverage more than 95%).
One of the main consequences of using aggregate estimations is bilateral
asymmetries that will occur as the statistical estimations are not based on a
country-by-country data.
II – Bilateral level estimations
In the cases where there is reported data on a bilateral basis by LBS
reporting country vis-à-vis the domestic country, mirror data exercises are
more effective and precise. The term ’domestic country’ used below refers to
the country that we wish to estimate households’ cross-border deposits. The
following three scenarios are defined according to the available information
17
for counterparty sectors N, F, P and H :
1. Deposits placed by households (H) – liabilities of banks vis-à-vis
sectors H and N are reported:
In this first scenario, households’ deposits abroad will correspond to the
bilateral deposit liabilities of the counterpart reporting country to the
domestic country. Such data available only from Q4 2013. If both sectors P
and H are not reported, bilateral positions can be estimated backwards by
applying in each quarter the weighted average of the reported households
18
sector (H) in the sector N to the non-banks sector (N) positions.
2. Deposits placed by households (H) – liabilities of banks vis-à-vis P
amounts are reported prior to sector H:
In this second case, subsector H is reported at a later stage than sector P,
we propose to use information of H when available and estimate H positions
19
for each of the non-reported quarter using, the average weight of the
households sector (H) in the total amount of sector P (from the quarters with
reported data on sector H), which is more precise than using sector N
mentioned above.

amounts for all previous quarters when none of the sub-sectors of P are available. An improved
alternative is to use moving average or average from the latest 4 or 8 quarters and apply the
share to reported sector N amounts.

17 See footnote 13. We propose to apply the same method for individual reporting countries to
get better estimates not only for bilateral positions but also for total of ‘all reporting countries’
when re-aggregated.
18 Which can be a simple / weighted or moving average.
19 Which can be a simple / weighted or moving average.
420 | I S I W S C 2 0 1 9