(AGENPARL) - Roma, 21 Febbraio 2025 - (AGENPARL) – ven 21 febbraio 2025 Notes on
Financial
Stability and
Supervision
No. 44
January 2025
Table of contents
Overview………………………………….1
Introduction
and main conclusions ……………….2
1. Insights on CRE loans
and collateral………………………..3
2. Collateral valuations
according to AnaCredit…………… 5
3. The impact of collateral
revaluation on the current
loan-to-value ratio………………..8
3.1 LTV-C ……………………………8
3.2 Simulation analysis ………10
References………………………………14
Annex A: Technical notes…………15
Annex B: Additional Figures
and Tables………………..16
The papers published in the
Notes on Financial Stability
and Supervision series
reflect the views of the authors
and do not involve
the responsibility
of the Bank of Italy
Designed by
the Printing and Publishing Division of the Bank of Italy
ISSN 2284-4198 (online)
Banks’ collateral valuation for
commercial real estate loans:
lessons from AnaCredit
Federica Ciocchetta – Raffaella Pico – Ivan Quaglia*
Overview
This note presents an analysis of collateralized CRE loans
extended by Italian banks to non-financial corporations,
using the granular information reported in AnaCredit. First, it
provides insights into how frequently banks update real estate
collateral values, by comparing their reported valuations over
time with changes in market prices. It then proposes simple
simulation exercises to assess the impact of hypothetical
declines in collateral prices on loan-to-value ratios and on
expected losses on banks’ CRE portfolios.
The findings of the analysis indicate that the value of real
estate collateral for CRE loans is frequently revalued. While
the revaluations do not seem to be fully in line with changes in
average market prices, deviations are contained and might be
influenced by the fact that aggregate market prices do not fully
capture the heterogeneity of real estate characteristics. The
results of the simulations show that the risks for the Italian
banking system associated with both baseline and (various)
adverse assumptions for real estate price developments
are generally contained. Indeed, the increase in expected
losses on CRE loans would remain moderate even in case of
a sharp decline in property prices (larger than that observed
historically) and a significant increase in the probability of
default of the non-financial firms to which the collateralized
CRE loans have been extended.
* Directorate General for Economics, Statistics and Research.
Introduction and main conclusions
As seen in many countries during the global financial crisis, adverse real estate
price developments can pose significant risks to financial stability and negatively
impact banks’ resilience (Agnello and Schuknecht, 2011; Claessens et al., 2009;
Mian and Sufi, 2014; Cerutti et al., 2015). Combined with a high borrowing costs
and a weak economic outlook, a decline in property prices could result in an increase
in borrowers’ defaults and credit losses for banks. In light of these developments,
there is a growing concern among policymakers and regulators about the potential
impact of a downturn in the real estate market, especially in the commercial real
estate sector (CRE), on the financial system and the broader economy.
This note presents an analysis of collateralized CRE loans, as defined by the
Recommendation ESRB/2019/03, granted to non-financial corporations (NFCs).
It is worth noting that collateral includes not only non-residential immovable
properties (such as offices, commercial premises and industrial plants), but also
residential ones, provided that they are associated with a loan granted to a legal
entity. To gain insights into the valuation of real collateral by banks in Italy, we
utilize granular information collected in AnaCredit, comparing banks’ reporting
of the real estate collateral valuations over time with observed changes in market
prices.
The value update policy following real estate price movements has an impact on
the denominator of the current loan-to-value ratio (LTV-C) associated with a given
contract. As the LTV-C is a key variable to assess the vulnerability of banks toward
the real estate sector, we then examine the distribution and breakdowns of this
indicator. Finally, we propose a simple simulation exercise to assess the impact of
expected and hypothetical drops in real estate prices on LTV-C and on expected
losses in banks’ CRE loan portfolios. The work focuses on the share of loans with
a LTV-C greater than 80 per cent (hereafter referred to as ‘high LTV-C loans’) as
these are associated with greater losses in the event of default and possibly to a
higher probability of default (Mokas and Nijskens, 2019).
The main results of the analysis are the following:
• AnaCredit data show that the value of real estate collateral are
frequently revalued. Between 2019 and 2023, the revaluations would have
affected approximately 70 per cent of the properties used as collateral each
year, a relatively high share in international comparison (Horan et al. 2023).
While the revaluations do not seem to be fully in line with changes in average
market prices, deviations are contained. In line with the mild downward trend
in prices in the non-residential real estate market and due to the predominance
of these properties in the collateral, over 50 per cent of the real estate collateral
experienced downward revaluation changes between 2019 and 2023. Deviations
from the price trend might be influenced by the fact that aggregate market
prices do not fully capture the heterogeneity of real estate characteristics.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
• In December 2023, the average LTV-C was 55 per cent, slightly lower
than the 60 per cent recorded in December 2019. The share of high LTV-C
loans over total CRE loans was about 13 per cent (11 per cent if we consider
only performing loans).
• Simulation exercises show that the risks for the Italian banking system
associated with both baseline and (various) adverse assumptions for
real estate price developments are generally contained: the share of
loans with a high LTV-C would increase to more than 16 per cent if we apply a
large and differentiated shock to the office and commercial property segment.
Indeed, the increase in expected losses on CRE loans would remain moderate
even in case of a sharp decline in property prices (larger than that observed
historically) and a significant increase in the probability of default of the
non?financial firms to which the collateralized CRE loans have been extended.
The rest of the note is structured as follows. Section 1 reports a description of
collateralized CRE loans. Section 2 offers insights into AnaCredit valuations and
section 3 presents the analysis of LTV-C and the results of the simulation exercise.
1. Insights on CRE loans and collateral
The analysis relies on AnaCredit, a harmonized and granular database with
detailed information on a sample of bank loans granted to legal entities. Our
dataset focuses on collateralized CRE loans to NFCs by Italian banks.1 The data
are analyzed according to different breakdowns, including: the type of property
(residential,2 offices and commercial premises, other commercial properties); the
economic sector of the borrower (construction firms, real estate companies, other
firms); the loan purpose; the geographical area where the property is located (for
further details see Annex A.1).
In December 2023 the total amount of collateralized CRE loans collected in
AnaCredit was about 13 per cent (68 billion of euros) of total loans to NFCs, with
38 billion of loans associated with offices and commercial premises (figure 1a).
About 30 per cent (20 billion euro) of collateralized CRE loans are backed by other
commercial collateral and only a small part (16 per cent, about 11 billion euro) by
residential (RRE) collateral.
A mere 15 percent of collateralized CRE loans are granted to construction firms,
about 30 per cent to real estate companies and the majority to firms of other
economic sectors (figure 1b). CRE loans are mainly collateralized by commercial
property (e.g. offices, commercial premises like shops, and other commercial
The CRE loans collected in AnaCredit are about the 75 per cent of CRE loans reported in Finrep, for both collateralized
and total CRE loans.
In line with the ESRB definition, all real estate protections in AnaCredit can be considered CRE if they are owned by legal
entities and not by households, in which case the residential properties should be classified as CRE, type residential. An
example is a dwelling property under development and pledged as collateral by a construction firm.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
increase in the probability of default of the non-financial firms to which the collateralized CRE
loans have been extended.
The rest of the note is structured as follows. Section 1 reports a description of collateralized CRE
loans. Section 2 offers insights into AnaCredit valuations and section 3 presents the analysis of LTVproperties like industrial plants) but for construction firms the share of residential
C and the results of the simulation exercise.
property is slightly higher than for the other two borrower categories. Commercial
properties are the main collateral also of loans granted for the purpose of financing
1 Insights
on CRE loans and collateral
properties ‘under development’, which do not exist at the contract origination
The analysis relies on AnaCredit, a harmonized and granular database with detailed information on
figure
1c), Our
which
represent
small share CRE
of the
a sample(‘construction
of bank loans investment’;
granted to legal
entities.
dataset
focuses
on acollateralized
loans to
total.
largest
share
estate
collateral
located
North
while
offices
NFCs by Italian banks. The data are analyzed according to different breakdowns, including: the type
and commercial
are commercial
the largest premises,
part of real
estate
collateralproperties);
in all the the
offices and
other
commercial
of property
(residential, 2premises
geographical
areas
(figure(construction
economic
sector of the
borrower
firms, real estate companies, other firms); the loan
purpose; the geographical area where the property is located (for further details see Annex A.1).
Figure
characteristicsof
ofCRE
CREloans
loans
Figure
1: 1:
characteristics
(billions of euros for panel a; per cent for panels b-d; December 2023)
(billions of euros for panel a; per cent for panels b-d; December 2023)
Type of property
b) Economic sector
Loan purpose: construction investment
d) Geographical area
Source: AnaCredit. Note: Only CRE loans collateralized by a real estate property. In the case a CRE loan is associated with
Note: Only CRE
collateralized
a realto
estate
In thethe
caselargest
a CRE loan
is associated
onecategories
more thanSource:
one AnaCredit.
type of collateral,
weloans
assigned
the by
theproperty.
type with
value.
Panelwith
b:more
the than
three
of collateral, we assigned the loan to the type with the largest value. Panel b: the three categories have been identified by using the ATECO
‘Construction’
classified
under
ATECO 41, 42, 43;
‘Real estate’
under ATECOclassified
68 (‘Rental and
management
of owned
have beencode.
identified
by includes
using enterprises
the ATECO
code.
‘Construction’
includes
enterprises
under
ATECO
41, 42, 43;
or leased properties’); ’Others’ the remaining sectors.
‘Real estate’ under ATECO 68 (‘Rental and management of owned or leased properties’);’Others’ the remaining sectors.
The CRE loans collected in AnaCredit are about the 75 per cent of CRE loans reported in Finrep, for both collateralized
in figure 2, there is certain heterogeneity of CRE exposures across banks.
and total As
CREillustrated
loans.
2 In line with
In December
2023, on
average
6.5 per cent
of bank can
total
was collateralized
by by
the ESRB definition,
all real
estatea protections
in AnaCredit
beloans
considered
CRE if they are owned
legal entities
households,
which
residential
properties
should
classified
residential.
a real estate property; approximately 50 per cent of all banks had a CRE loan share of
An example is a dwelling property under development and pledged as collateral by a construction firm.
less than 10 per cent of total loans, which corresponds to approximately 80 per cent
of total banking system assets. Only 10 per cent of all banks have a share of CRE loans
exceeding 20 per cent.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
which corresponds to approximately 80 per cent of total banking system assets. Only 10 per c
all banks have a share of CRE loans exceeding 20 per cent.
Figure
2: CRE
exposure
bybank
Figure
2: CRE
exposure by
of euros
andnumbers;
numbers; December
2023) 2023)
(billions(billions
of euros
December
1,200
1,000
5%-10% 10%-15% 15%-20% 20-25% 25%-30% 30%-35% 35-40%
Share of CRE loans over total
Banks (lhs)
Assets
Source: Finrep and
report.
a solo basis.
Source:supervisory
Finrep and supervisory
report.
Thedata
data areare
on a on
solo basis.
Collateral
valuations according to AnaCredit
2. Collateral valuations according to AnaCredit
In this section,
we investigate
changes in
banks’invaluation
with respect
to the
of real e
In this
section, we investigate
changes
banks’ valuation
with respect
todynamics
the dynamics
estate pricehow
indices
to investigate
howcollateral
banks update
the collateral
reported
price indicesoftoreal
investigate
banks
update the
values
reportedvalues
in AnaCredit.
in AnaCredit.
In accordance with EU regulations and supervisory guidelines banks are required to
update the collateral value periodically. The Capital Requirements Regulation (CRR)
art. 208 regulates requirements for immovable property collateral to be eligible for the
risk mitigation. Specifically, some requirements regard the monitoring of collateral
values over time and the valuation approaches. Institutions have to monitor the value
of property on a regular basis and, at a minimum, once every year for commercial
immovable property and once every three years for residential property. Moreover, the
property valuation is reviewed when information available to institutions indicates
that the value of the property may have declined materially relative to general market
prices (for details see Annex A.2). For the analysis of property valuation, we consider
the variable ‘protection value’ (PV) reported in AnaCredit, which is the value of the
real estate property, generally the market value assessed by the bank and updated
periodically. This value should change over time reflecting the changes in market prices.
Concerning the price trends, since 2019 house prices in Italy have consistently
increased, although less than in the rest of Europe (figures 3a and 3b). Conversely, nonresidential property prices have been declining for at least a decade although with some
heterogeneity: office and retail properties decreased more than the industrial segment.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
Concerning the price trends, since 2019 house prices in Italy have consistently increased, although
less than in the rest of Europe (figure 3a and b). Conversely, non-residential property prices have
been declining for at least a decade although with some heterogeneity: office and retail properties
decreased more than the industrial segment.
Figure
estate
price
developments
Figure
estate
price
developments
(Indices 2019 =100)
(Indices 2019 =100)
Residential real estate
2023-Q4
2023-Q3
2023-Q2
2023-Q1
2022-Q4
2022-Q3
2022-Q2
2022-Q1
2021-Q4
2021-Q3
2021-Q2
2021-Q1
2020-Q4
2020-Q3
2020-Q2
2020-Q1
2019-Q4
2019-Q3
2019-Q2
2024-Q1
2023-Q4
2023-Q3
2023-Q2
2023-Q1
2022-Q4
2022-Q3
2022-Q2
2022-Q1
2021-Q4
2021-Q3
2021-Q2
2021-Q1
2020-Q4
2020-Q3
2020-Q2
2020-Q1
2019-Q4
2019-Q3
2019-Q2
2019-Q1
2019-Q1
b) Non-residential real estate
Source: ECB SDW, OECD. Note: For non-residential properties, the indices are based on non-harmonized data and they
SDW, OECD. Note: For non-residential properties, the indices are based on non-harmonized data and they are not fully
areSource:
not fully
comparable.
comparable.
Figure 4a summarizes the change in valuation of AnaCredit collateral at the system level. In line with
4a summarizes
change
in non-residential
valuation of AnaCredit
at due
the system
theFigure
mild downward
trend inthe
prices
in the
real estatecollateral
market and
to the
level.
downward
trend
prices
non-residential
estate
predominance of these properties in the collateral, over 50 per cent of the real estate collateral
market and
due to the
predominance
these properties
the Overall,
collateral,
50 per
experienced
downward
revaluation
changesofbetween
2019 and in
2023.
theover
change
valuations
has collateral
involved about
70 per cent
of collateral
in AnaCredit
(figure 4b);
cent of for
theeach
realyear
estate
experienced
downward
revaluation
changes
between
percentage
comparison
other
countries
(Horan
2023).
2019 and 2023. Overall, the change in valuations for each year has involved about 70
cent ofiscollateral
AnaCredit
4b); this percentage
in the
Theper
evidence
confirmedin
we look (figure
at the breakdowns
(figure B.1 inisthe
Annex
B):comparison
the median
with other
EUfrom
countries
(Horan
et December
al., 2023).2023 remains consistently around minus 1 per
valuation
change
December
2019 to
cent, with some negative peaks, particularly in the construction sector (2 per cent decline).
Conversely, there are observable instances of upward revaluation changes in both the East and West
Figure4:4:Collateral
Collateralvaluation
valuation
areas of the Northern part of Italy. Figure
a) Distribution for Italian banking system
(variation of December 2023 on December 2019)
Revalued real estate collateral in Italy
Source: AnaCredit. Note: We calculate the valuation change at the level of single collateral/instrument; the charts report
Note:
We calculateinthe
valuation
change atclass
the level
of whose
single collateral/instrument;
charts
report
the share
of collateral
the Source:
share AnaCredit.
of collateral
properties
revaluation
(i.e.
valuation changethe
given
interval).
exclude
properties in each revaluation class (i.e. whose valuation change is a given interval). We exclude NPL from the analysis.
NPL from the analysis.
detail, figure
5 compares
average
market
price
dynamics with
changes
banks’
valuation,
Theinevidence
is confirmed
we look
at the
breakdowns
(figure
B.1 inin
Annex
B): the
plotting growth rates in real estate property price index against value adjustments of collateral
median valuation change from December 2019 to December 2023 remains consistently
items, by different sub-segments (fig. 5, panels a, b, c), both on year-o-year basis and over the period
around
minus 1 to
perthe
cent,
with some
negative
peaks, particularly
in the construction
sector
2019-23.
According
AnaCredit
report,
banks update
real estate property
valuations, although
the change is generally lower compared to change in the property price index. In the case of the
residential segment all changes in valuations are below that of the property price index; in the case
of the commercial segment, on the other hand, it6 appears that the value of properties pledged as
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
collateral for loans has partially reflected the depreciation recorded in the market on average.
(2 per cent decline). Conversely, there are observable instances of upward revaluation
changes in both the East and West areas of the Northern part of Italy.
More in detail, figure 5 compares average market price dynamics with changes in
banks’ valuation, plotting growth rates in real estate property price index against
value adjustments of collateral items, by different sub-segments (fig. 5, panels a, b, c),
both on year-o-year basis and over the period 2019-23. According to the AnaCredit
report, banks update real estate property valuations, although the change is generally
lower compared to change in the property price index. In the case of the residential
segment all changes in valuations are below that of the property price index; in
the case of the commercial segment, on the other hand, it appears that the value
of properties pledged as collateral for loans has partially reflected the depreciation
recorded in the market on average. Similar results can be observed for the change
over the period of 2019-23 (fig. 5d).
Figure 5: real estate price indices and revaluations by type of property
Figure 5: real estate price indices(Per
revaluations by type of property
cent)
(Per cent)
Residential properties
(annual change)
b) Offices and commercial premises
(annual change)
Housing prices
Change valuation (p10)
Change valuation (median)
Change valuation (p90)
Change valuation (average)
Change valuation (q1)
Change valuation (q3)
Other commercial properties
(annual change)
d) All sectors
(change 2023-19)
Industrial property prices
Change valuation (p10)
Change valuation (median)
Change valuation (p90)
Change valuation (average)
Change valuation (q1)
Change valuation (q3)
1 quartile
Offices and commercial
premises
median
3 quartile
Other CRE
Growth rate of property prices
Source: OMI, AnaCredit – Note: We calculate the series of price indexes for offices and commercial premises as the
Source: OMI,
AnaCredit.
We two
calculate
the seriesrespectively.
of price indexes
as the
average
theindustrial
process of
average
of the
processNote:
of the
segments,
Infor
theoffices
caseand
of commercial
other CRE,premises
the series
refers
toofthe
the two segments,
respectively.
In the caseonly
of other
CRE, the
series
refers to the industrial
segment. We consider the collateral only for loans
segment.
consider
collateral
loans
performing
status.
in the performing status.
The differences
impact of collateral
on the current
ratio
betweenrevaluation
AnaCredit revaluation
andloan-to-value
property price
indexes may
depend on a number of factors. First, we use a measure for the price change at the
3.1 LTV-C
The LTV-C is a key variable to assess the vulnerability of banks toward the real estate sector: a high
7 the event of default and, ex-ante, with a higher
value of LTV-C is associated with higher losses in
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
probability of default (Mokas and Nijskens, 2019).
aggregate
level that might not capture well the valuation dynamics of the three
groups of collateral identified. Second, at the granular level, collateral value reflects
2020characteristics
specific
as whether
the property is an old building, a new one
Housing prices
Change valuation (average)
Change valuation
(p10)
Change
valuation (q1)
or under
construction,
property
location or the quality of the building, for
Change valuation (median)
Change valuation (q3)
Change valuation (p90)
instance in terms of energy efficiency class; this can be especially true for the RRE
c) Other
commercial
propertiesfor CRE loans. Third, the assessment
d) All sectors
properties
as collateral
of the market
(annual change)
(change 2023-19)
value can be based on long-term considerations that keep the value constant for
a2 while. In the case the loan is in the performing status, a common bank practice
updating collateral valuation is to apply statistical approaches; valuations are
reviewed
only in the case of significant difference relative to general market prices.
Overall,
the updates of the collateral values-5 reported in AnaCredit seem to reflect
price developments over time. In this sense,
collateral-based indicators (such as
LTVs),
constructed using these data, are consistent
with price index changes and are
Offices and commercial
Other CRE
premises
appropriate
forpricesanalyzing Change
loans
Industrial property
valuation
(average) (see Annex A.3 and Chart B.2).
Change valuation (p10)
Change valuation (median)
Change valuation (p90)
Change valuation (q1)
Change valuation (q3)
1 quartile
median
3 quartile
Growth rate of property prices
Source: OMI, AnaCredit – Note: We calculate the series of price indexes for offices and commercial premises as the
3. The
impact
ofthe
collateral
revaluation
onthethe
loan-to-value
average
of the
process of
two segments,
respectively. In
casecurrent
of other CRE,
the series refers ratio
to the industrial
segment. We consider the collateral only for loans in the performing status.
3.1 LTV-C
The impact
LTV-C is
key variable
to assesson
of banks toward
3 The
of acollateral
revaluation
thevulnerability
current loan-to-value
ratio the real
estate sector: a high value of LTV-C is associated with higher losses in the event
3.1ofLTV-C
default and, ex-ante, with a higher probability of default (Mokas and Nijskens,
2019).
LTV-C is a key variable to assess the vulnerability of banks toward the real estate sector: a high
value of LTV-C is associated with higher losses in the event of default and, ex-ante, with a higher
The bank policy for collateral value update following real estate price movements
probability of default (Mokas and Nijskens, 2019).
affects the denominator of this indicator, which also changes at the numerator for
policy forofcollateral
update
following real
price movements
affects
thebank
repayment
the debt.value
Based
on AnaCredit
data,estate
it is possible
to calculate
thethe
denominator
thisloan
indicator,
which also
changes at theand
numerator
for the repayment of the debt.
LTV-C for of
to explore
heterogeneities
risk concentration.
Based on AnaCredit data, it is possible to calculate the LTV-C for each loan to explore
heterogeneities
and risk date
concentration.
At a given reference
the LTV-C at the loan level is:
At a given reference date the LTV-C at the loan level is:
????????????????????,????
????????????????????,???? = ????
??????? ????????????????????????,????,????
where loani,t is the residual amount of the loan i granted by a given bank at time t and
at the denominator we consider the sum of all properties values j associated to the loan
i, at the current value reported in t.
In AnaCredit there are two possible attributes for the collateral value: the protection
value (PV) and the protection allocated value (PAV, see figure B.3 in the Annex B). PV is
the monetary value of the protection item that was established at the latest valuation
date prior to the reporting reference date; it reflects the total value of the protection.
PAV is instead the maximum amount of the PV available as credit protection for the
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
where loani,t is the residual amount of the loan i granted by a given bank at time t and at the
denominator we consider the sum of all properties values j associated to the loan i, at the current
value reported in t.
In AnaCredit there are two possible attributes for the collateral value: the protection value (PV) and
theinstrument
protection allocated
value
(PAV,
see figure
in the Annexconsiderations.
B). PV is the monetary
value of the
based on
banks’
internal
riskB.3
management
In the following,
protection
established
latest
valuation
prior
reporting
reference
we consider the LTV-C with PAV, as similar to the approach used by banks and
date;
reflects
total
value
protection.
instead
maximum
amount
the PV
conservative with respect to use the (market) property value.
available as credit protection for the instrument based on banks’ internal risk management
considerations.
In2023,
the following,
we consider
the LTV-C
similar to the
approach
In December
the average
LTV-C was
55 perwith
cent,
in aasdecreasing
trend
from the
by banks
conservative
witharespect
use the (market)
value.across different
60 perand
in 2019
(figure 6a);
certaintoheterogeneity
canproperty
be observed
In December
2023,
the average
LTV-C
wasThe
55 LTV-C
per cent,
a decreasing
trend fromfirms
the 60and
perfor
breakdowns
(figure
6, panels
b-d).
is in
higher
for construction
in 2019
(figure
certain
heterogeneity
observed
across
different
breakdowns
(figure
loans whose purpose is construction investment. By geographical area, the LTV-C is6,
panels
b-d).for
LTV-C is higher
construction firms and for loans whose purpose is construction
higher
properties
in thefor
North-East.
investment. By geographical area, the LTV-C is higher for properties in the North-East.
Figure
Figure6:
6: LTV-C:
LTV-C:distribution
distributionand
andbreakdowns
breakdowns
Source:
AnaCredit
Note:
included
a. We
calculate
LTV at the
by allocated
using the
Source:
AnaCreditdata.
data. Note:
are are
included
only inonly
panelin
a. panel
We calculate
at the instrument
levelsinstrument
by using the levels
protection
value at the
denominator.
Panel
a: thedenominator.
red point indicated
the weighted
average
of LTV-C
by loanthe
amount.
The same
chart forof
performing
protection
allocated
value
at the
Panel
a: the red
point
indicated
weighted
average
LTV-C
by loan
loans is reported in the Annex B (figure B.4). Panels b-d: each figure represents the distribution of LTV in December 2023, total and for each
amount.
chart
performing
loans
reported
Annex
(figure
B.4).
Panels
figure
value of the variable of interest (construction investment, borrower type and geographical area). The central line of each box-plot represents
the median
value,
the bottom and
upper lines
of theand
box the
and 3rd
quartile.
The values
outside
inter-quantile
interval areinvestment,
not shown.
represents
distribution
of LTV-C,
total
for1st
value
of the
variable
of the
interest
(construction
borrower type and geographical area). The central line of each box-plot represents the median value, the bottom and
upper lines of the box the 1st and 3rd quartile. The values outside the inter-quantile interval are not shown.
The distribution of LTV-C varies across banks (figure 7). The LTV-C are higher for
cooperative banks and lower for other categories. In terms of distribution of the
indicator, we should pay particular attention to high LTV-C loans as they are generally
3 AnaCredit states that, in order to determine a protection allocated value, an intermediary follows the collateral
considered riskier and can be related to greater losses in case of defaults, especially in
allocation principles used internally for risk management purposes. In this respect, AnaCredit does not provide
caseallocation
of a strong
dropfor
in determining
property valuations.
any the
standard
algorithm
the protection allocated value.
AnaCredit states that, in order to determine a protection allocated value, an intermediary follows the collateral allocation
principles used internally for risk management purposes. In this respect, AnaCredit does not provide any standard
allocation algorithm for determining the protection allocated value.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
distribution of LTV-C varies across banks (figure
7).AnaCredit.
The LTV-CNote:
are higher
cooperative
Source:
LTV is for
calculated
at thebanks
instrument level by using the protecti
denominator.
Theindicator,
four bank we
categories
1) Significant banks; 2) BCC (including ba
lower for other categories. In terms of distribution
of the
shouldinclude:
pay particular
cooperative
groups
– ICCREA
and other
ntion to high LTV-C loans as they are generally
considered
riskier
andand
canCCB
be-related
tocooperative
greater banks not belonging to them); 3
banks). The central line of each box-plot represents the median value, the bottom and upper
es in case of defaults, especially in the case of a strong drop in property valuations.
3rd quartile. The values outside the inter-quantile interval are not shown.
Figure 7: LTV-C by bank category
Figure 7: LTV-C
by bank
(December
2023) category
(December 2023)
Figure 8: LTV-C greater than 80 per cent
by property type
Figure 8: LTV-C greater
80 per
cent by property typ
(Per cent;
December
2023)
(Per cent; December 2023)
ce: AnaCredit. Note: LTV is calculated at the instrument
by using
the we
protection
allocated
valueinatterms
the of the type of collateral. We cons
Source:level
AnaCredit.
Note:
distinguish
CRE loans
Source: AnaCredit.
LTVcorresponding
is calculated
at the
level
ominator. The four bank categories
include:Note:
1) Significant
banks;
2)toinstrument
(including
banks
belonging
Source:
AnaCredit.
Note:
distinguish
CRE loans inSee
terms
of theB.5 for r
the case of different types of property used
as collateral.
figure
using the protection allocated value at the denominator. The four
of collateral. cooperative
We consider also a mixed category corresponding
erative groups – ICCREA and by
other
cooperative
banks
belonging
them);
(excluding
bank categories include: 1) Significant banks; 2) BCC (including
to the case of different types of property used as collateral. See
banks belonging
to the
cooperative
– ICCREA
s). The central line of each box-plot
represents
thetwo
median
value,groups
the bottom
andand
upper lines
thein1st
figureof
B.5the
for box
results
terms
of amount.
In December
2023,
CCB – and other cooperative banks
not belonging to
them); the
3) LSIshare of high LTV-C loans is 12 per cent of the overa
quartile. The values outside the
inter-quantile
interval
shown.
(excluding cooperative banks). The central line of each box-plot
Figure
to 10 billion, figure
represents the median value,aggregate
the bottom andlevel
upper (corresponding
lines of the
box the 1st and 3rd quartile. with
The values
outside
inter-quantile
different types of collateral. Banks with a high
interval
are not
shown. than 80 per cent by property type
LTV-C
greater
a low share
(Per cent; December
2023)of
CRE exposure in their loan portfolio.
8); a similar share can be
share of LTV-C above 80 p
In December 2023, the share of high LTV-C loans is 12 per cent of the overall CRE portfolio
at the aggregate level (corresponding to 10 billion, figure 8); a similar share can be observed
across loans with different types of collateral. Banks with a high share
of LTV-C above 80 per
cent have generally a low share of CRE exposure in their loan portfolio.
3.2 Simulation analysis
In this section, we describe a static simulation exercise to assess the impact of a
decline in real estate values on the LTV-C distribution, in particular on the share of
high LTV-C loans, and consequently on the estimation of the LGD and expected losses
related to CRE loan portfolios, for the whole banking system. We estimate the impact
underCRE
different
hypotheses
the property
valuation
developments;
we consider the
ce: AnaCredit. Note: we distinguish
loans in terms
of the type on
of collateral.
We consider
also a mixed
category
esponding to the case of different
types ofassumptions
property used as collateral.
See figure
B.5 for
results in terms
of amount.
following
on property
prices
(‘scenarios’,
Table
1) ofBaseline:
theloans
shock
realcent
estate
prices
consists
of a small
ecember 2023, the share
high LTV-C
is 12toper
of the
overall
CRE portfolio
at theincrease in the house
regate level (correspondingprices
to 10 billion,
a similar
can be observed
across loans
and a figure
slight 8);
decline
for share
non-residential
real estate
prices.
h different types of collateral. Banks with a high share of LTV-C above 80 per cent have generally
adverse:
a shock to real estate prices of -5 pp, similar to the highest annual
w share of CRE exposure2)
in Mildly
their loan
portfolio.
decrease of property prices observed in the past (over the period 2004q1-2023q4).
The magnitude of the shocks is defined taking into account historical values of property price (over the period 2004q12023q4) and Banca d’Italia forecasts of house prices. For the baseline scenario, in the case of non-residential properties
no forecasts are available and we assume a slight negative growth rate of prices, in line with the latest observations.
Scenario 4 is high adverse and never realized in the past. A higher drop for commercial properties prices (-15 per cent
over the years 2023-2024) are presented in Martini et al. 2023.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
3.2 Simulation analysis
In this section, we describe a static simulation exercise to assess the impact of a decline in real estate
values on the LTV-C distribution, in particular on the share of high LTV-C loans, and consequently on
the estimation of the LGD and expected losses related to CRE loan portfolios, for the whole banking
3) Differentiated shock: the shock is greater for offices and commercial premises (-15
system. We estimate the impact under different hypotheses on the property valuation
per cent),
isfollowing
limited for
other commercial
properties
(-2 per cent)
developments;
wewhereas
consider it
assumptions
on property
prices (‘scenarios’,
Table and
1 4):
high for residential properties (-5 per cent).
1) Baseline: the shock to real estate prices consists of a small increase in the house prices and a
4) slight
Highly
adverse:
the shock on
estate
valuation is of -10 per cent, twice as
decline
for non-residential
estate
prices.
large adverse:
as the maximum
price
decline
observed
period of
2) Mildly
a shock to annual
real estate
prices
of -5 pp,
similar historically
to the highest(over
annual
decrease
2004q1-2023q4).
property
prices observed in the past (over the period 2004q1-2023q4).
3) Differentiated shock: the shock is greater for offices and commercial premises (-15 per cent),
whereas it is limited for other commercial properties (-2 per cent) and high for residential
Table
1: summary of different scenarios of price shocks
properties (-5 per
cent).
4)Scenario
Highly adverse: the shock onPrices
real estate valuation is of -10Property
per cent,
twice as large as the
categories
maximum annual price decline observed historically (over the period 2004q1-2023q4).
1. Baseline
+1 per cent for residential and -0.5
Differentiated shock depending
-5 per cent
Same shock to all properties
per centof
fordifferent
other properties
on property
Table 1: summary
scenarios of price
shocks
2. Mildly adverse
Scenario
Prices
Property categories
Differentiated shock +1 per cent for
-5 per
cent for residential,
-15 for Differentiated
Differentiated
shock depending
1.3.Baseline
residential
and -0.5
shock
depending on
offices and commercial premises,
on property type
per cent for other
property type
-2 for properties.
other commercial
2. Mildly adverse
-5 per cent
Same shock to all properties
4. Highly adverse
-10 per cent
Same shock to all properties
3. Differentiated
-5 per cent for residential, -15 for
Differentiated shock depending on
shock
offices and commercial premises, -2
property type
for other commercial.
starting
point-10
of per
thecent
exercise is the stock of all outstanding
Highly
adverse
Same shock to performing
all properties loans as
of December 2023 and with a residual maturity greater than 1 year. The exercise is
i.e.point
the stock
loans remains
fixedofand
in the performing
status
throughout
Thestatic,
starting
of theofexercise
is the stock
all outstanding
performing
loans
as of December
analysis
period.
In addition,
do not
amortize
the loan
in the
LTV numerator.
and with
a residual
maturitywe
greater
1 year. The
exercise
is static,
i.e. the stock of loans
remains fixed and in the performing status throughout the analysis period. In addition, we do not
To assess
on numerator.
the banks’ balance sheets, we calculate the expected losses (EL)
amortize
the the
loanimpact
in the LTV
by using the following formula:
To assess the impact on the banks’ balance sheets, we calculate the expected losses (EL) by using
the following formula:
???????? = ???????? ? ???????????????? ? ????????????
where PD is the probability of default, LGDs is the loss given default (for the scenario s) and EAD is
the exposure at default. The PD value is the one reported in AnaCredit for loans in the IRB portfolio,
where
the probability
of default,
the isloss
given
default
scenario
whichPD
theisinformation
is available,
in theLGDs
other is
cases
equal
to the
median
PDthe
value.
s) and EAD is the exposure at default. The PD value is the one reported in AnaCredit
for loans in the IRB portfolio, for which the information is available, in the other cases
is equal to the median PD value.5
The magnitude of the shocks is defined taking into account historical values of property price (over the period
We start with
the PD
is constant
in allscenario,
the real
2004q1-2023q4)
and the
Bancaassumption
d’Italia forecasts
of house
prices.
For the baseline
, in estate
the caseprice
of nonscenarios;
we also
consider
additional
exercises
based
on different
assumptions
residential
properties
no forecasts
aretwo
available
and we assume
a slight
negative
growth rate
of prices, in lineon
the PDs:
latest 1)
observations.
Scenario
4 is high increase
adverse and
realized
in the
past. A higher
drop for
the case of
a moderate
innever
the PD
for the
baseline
scenario
(+2commercial
pp for
properties
prices in
(-15the
per construction
cent over the years
2023-2024)
are presented
in Martini etand
al. 2023.
companies
sector
and real
estate companies
+1 pp for other
economic
a large increase for the highly adverse scenario (+4 pp and +2 pp,
median of sectors);
PDs is 2 per2)cent.
respectively). The EAD is given by the 10
outstanding value of the loan in December 2023.
The median of PDs is 2 per cent.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
consider two additional exercises based on different assumptions on PDs: 1) the case of a moderate
increase in the PD for the baseline scenario (+2 pp for companies in the construction sector and real
estate companies and +1 pp for other economic sectors); 2) a large increase for the highly adverse
scenario (+4 pp and +2 pp, respectively). The EAD is given by the outstanding value of the loan in
The LGD 2023.
under scenario s is estimated as a function of LTV-C (which depends on property
December
prices), administrative costs related to the litigation procedure and a discounting factor
LGD under
s isfollowing
estimated formula
as a function
of LTV-C
(which
prices),
(Discount),
byscenario
using the
(similar
to the
one depends
proposedoninproperty
Gross, and
We start
assumption
constant
estate
price
scenarios;
we also
administrative
costs related
to the
litigation
Población García,
2017 and
Jur?a
et al, procedure
2020): and a discounting factor (Discount), by using
consider
additional
exercises
PDs: 1) the
case2017
of aand
moderate
thetwo
following
formula
(similarbased
to the on
onedifferent
proposedassumptions
in Gross, andon
Población
García,
Jur?a
increase
in 2020):
the PD for the baseline scenario (+2 pp for companies in the construction sector and real
et al,
estate companies and +1 pp for other economic sectors); 2)1a large increase for the highly adverse
= ???????????????????? + ???????????? ?0, ?1 ?
? ????????????????????????????????? ?
scenario (+4 pp and +2 pp,????????????????
respectively).
The EAD is given
by the
outstanding value of the loan in
????????????????????
December 2023.
The objective of our exercise is to determine the variation in LGD due to the change in the collateral
valuation
consequently,
in the LTV-C
under different
scenarios.
The LGD
underand,
scenario
s is estimated
as a function
of LTV-C
(which depends on property prices),
The objective of our exercise is to determine the variation in LGD due to the change in
administrative
costs
related
a discounting
factor
(Discount),
estimate
the to
impact
these procedure
scenarios
banks’
sheets
in termsbyofusing
thethen
collateral
valuation
and,litigation
consequently,
in the
LTV-C
underbalance
different
scenarios.
the following
formula
(similar
proposed
Gross,
Población
García,
Jur?a
variation. For this purpose, we assume that the administrative costs and the discount rate are
et al, 2020):
We then6 Note
estimate
of these
scenarios on
balanceofsheets
in terms
that the
the impact
results may
underestimate
thethe
ELsbanks’
for a number
reasons,
for instance,
constant.
1 effects, such
ELsimulation
variation. For
weaccount
assumesecond-round
that the administrative
costs
the discount
doesthis
notpurpose,
take into
as and
a massive
sell-off of
????????????????
= ???????????????????? + ???????????? ?0, ?1 ?
? ????????????????????????????????? ?
collateral,
may cause
to fall
further.
rate are which
constant.
Note prices
that the
results
may underestimate
the ELs for a number of
????????????????????
reasons, for instance, the simulation does not take into account second-round effects,
Figure 9:of
simulation
exercise
The objective
our exercise
is to determine the variation in LGD due to the change in the collateral
such as a massive sell-off of collateral, which
may cause prices to fall further.
(Per cent) scenarios.
valuation and, consequently, in the LTV-C under different
a) Distribution of LTV-C
b) Expected losses
The impact of a shock on real estate valuation, and hence on the LTV-C, is limited in
We then
estimate
impact and
of these
scenarios
banks’
sheets(figure
in terms
almost
all the
scenarios
moderate
underon
thethe
severebalance
assumptions
9a).of EL
variation.
For thiswith
purpose,
we observed
assume that
the administrative
andscenario
the discount
rate are
Compared
the value
in December
2023, in thecosts
baseline
the share
constant.
NoteLTV-C
that the
results
underestimate
thethe
ELsfigure)
for a number
of reasons,
instance,
of high
loans
(green
yellow parts in
for performing
loansfor
would
the simulation
does not
take into
account
second-round
effects, adjustments
such as a massive
sell-off of
remain almost
constant
to about
11 per
cent, as the negative
of valuation
collateral, which may cause prices to fall further.
0.08%
0.07%
0.06%
0.04%
0.03%
0.02%
Figure 9: simulation exercise
Titolo asse
0.05%
0.01%
Figure 9: simulation exercise
1. Baseline
2. Mild
3. Diff shock 4. High adv 1b. Baseline
4b. High
(Per cent)
with PD
adverse with
increase
PD increase
Distribution of LTV-C
b) Expected ELlosses
EL change over RWA (rhs)
change – billion (lhs)
0.00%
0.08%
0.04%
Titolo asse
Source: AnaCredit, Corep (on individual-basis). Note: panel a only performing
loans are considered. The red point
0.07% is the
average LTV-C, weighted by loan amount. Panel b: the EL change0.70is calculated for each scenario with respect
0.06% to the
current EL (December 2023). Only performing loans are considered.
0.05%
The impact of a shock on real estate valuation, and hence
on the LTV-C, is limited in almost
0.03% all the
scenarios and moderate under the most severe assumptions
(figure 9a). Compared with 0.02%
the value
0.01% yellow
0.10 share of high LTV-C loans (green and
observed in December 2023, in the baseline scenario the
0.00%
parts in the figure) for performing loans would remain almost
constant
to about 11 per
cent,
as the
1. Baseline
2. Mild
3. Diff shock 4. High adv 1b. Baseline
4b. High
with PD
increase
EL change over RWA (rhs)
adverse with
PD increase
EL change – billion (lhs)
We calculate the discount factor by the following formula: 1/(1+rate)^T, where ‘rate’ is a risk-free interest rates
(such as 3-month Euribor, with a possible add-on) and T is the time of disposal. Based on the data in December
Source:
AnaCredit,
Corep (on
individual-basis).
Note: panel
a only performing
loans
are considered.
red point
about
Euribor
the duration of
litigation
procedure,
ifloans
we assume
administrative
costs
equal
to 30 is
Source:
AnaCredit,
Corepby
(onloan
individual-basis).
Note: Panel
a only
performing
are considered.
The redscenario
point is the
average
LTV-C,to the
average
LTV-C,
weighted
amount.
Panel
change
calculated
respect
of the
theb:average
LGD is
per cent,
a value
consistent
thecurrent
LGD for
corporate
weighted
byloan
loan amount,
amount. Panel
the EL change
is 32
calculated
for each
scenario
with respect
to the
EL the
(December
2023).portfolio
performing
loansapproach
are2023).
considered.
current
EL (December
performing
considered.
under
the IRB
reported
in Corep.loans
Someare
insights
for the calculation of LGD are reported in EBA 2017.
11 and hence on the LTV-C, is limited in almost all the
The impact of a shock on real estate valuation,
calculate
discount
factor
following
formula:
1/(1+rate)^T, where
‘rate’9a).
is a risk-free
interest with
rates (such
scenarios and moderate under the most severe assumptions
(figure
Compared
the value
as 3-month Euribor, with a possible add-on) and T is the time of disposal. Based on the data in December 2023 about
observed Euribor
in December
2023,ofinlitigation
the baseline
scenario
theadministrative
share of high
(green
and the duration
procedure,
if we assume
costs LTV-C
equal to loans
30 per cent
of theand
loanyellow
amount,
the average
LGD is 32 perloans
cent, a would
value consistent
the LGDconstant
for the corporate
portfolio
IRB as the
parts in the
figure)
for performing
remain
almost
to about
11under
per the
cent,
approach reported in Corep. Some insights for the calculation of LGD are reported in EBA 2017.
We calculate the discount factor by the following formula: 1/(1+rate)^T, where ‘rate’ is a risk-free interest rates
(such as 3-month Euribor, with a possible add-on) and T is the time of disposal. Based on the data in December
2023 about
we assume
BancaEuribor
d’Italia and the duration of litigation procedure, if Notes
on Financial administrative
Stability and Supervisioncosts
No. 44 -equal
Januaryto
202530 per
cent of the loan amount, the average LGD is 32 per cent, a value consistent with the LGD for the corporate portfolio
of non-residential properties offset the positive ones of residential properties. The share
would increase by 2 pp in the mildly adverse scenario and to more than 16 per cent in
adjustments
of valuation
non-residential
the adverse andnegative
differentiated
shock scenarios.
In theof
latter,
the weightedproperties
average of offse
residential
would
increase by2023.
2 pp The
in the
mildly
LTV-C rises to 60
per cent,properties.
from aboutThe
54share
per cent
in December
share
of adve
16 100
per per
in rises
the adverse
differentiated
shock scenarios.
loans with LTV-C
above
from 5 per
observed in December
2023 In
to 8 per cent under
sharpofdecline
in house
prices.
average
LTV-C rises
to 60
per cent, from about 54 per cent in December
with LTV-C above 100 per cent rises from 5 per cent observed in December
Moreover, the variation
of thein
ELhouse
underprices.
sharp decline
Figure 10: Simulation exercises – distribution
the different scenarios regarding the
of the share of high LTV-C loans by scenario
property prices isFigure
very limited
(figure exercises
10: Simulation
– distribution
of the share of high LTV(Per cent)
9b), with a value of about 70 million
(Per cent, share)
euros for all the banking system
(less than 0.01 per cent in terms of
RWA) in the case of the high adverse
scenario. Only if we also assume that
the PDS will change does the increase
in EL become moderate: it raises to
just above 800 million (about 0.07
per cent in terms of RWA) under the
high adverse scenario (scenario 4)
combined with a sharp increase of
PDs. Some heterogeneity across banks
can be observed both in terms of the
Source: AnaCredit data.
Source: AnaCredit data.
share of high LTV loans (figure 10) and
variation in estimated EL, depending
on the initial distribution
of LTV-C
andofonthe
theELheterogeneity
in thescenarios
property regarding
types
Moreover, the
variation
under the different
used as collateral
for CRE
loans.
limited
(figure
9b), with a value of about 70 million euros for all the bankin
per cent in terms of RWA) in the case of the high adverse scenario. Only if
PDS will change does the increase in EL become moderate: it raises to just
0.07 per cent in terms of RWA) under the high adverse scenario (scenario 4
increase of PDs. Some heterogeneity across banks can be observed both
high LTV loans (figure 10) and variation in estimated EL, depending on the i
C and on the heterogeneity in the property types used as collateral for CRE
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
References
Recommendation ESRB/2019/03, Recommendation on ‘closing real estate data gaps’,
emending Recommendation ESRB/2016/14.
ESRB Report (2023). Vulnerabilities in the EEA commercial real estate sector.
Agnello, L. and L. Schuknecht (2011). Booms and busts in housing markets: Determinants
and implications. Journal of Housing Economics, Vol. 20(3): 171-190.
Cerutti, E., J. Dagher and G. Dell’Ariccia (2015). Housing Finance and Real Estate Booms:
A Cross Country Perspective. IMF Staff Discussion Note.
Claessens, S., K.M. Ayhan and M.E. Terrones (2009). What Happens during Recessions,
Crunches and Busts? Economic Policy, Vol. 24(60): 653-700.
EBA 2017, Final Guidelines on PD and LGD estimation, Guidelines EBA-GL-2017-16.
Gross, M. and J. Población García (2017). Assessing the efficacy of borrower-based
macroprudential policy using an integrated micro-macro model for European households.
Economic Modelling, Vol. 61: 510-528.
Horan, A., B. Jarmulska and E. Ryan (2022). Understanding banks’ response to collateral
value shocks – insights from AnaCredit and the COVID-19 shock in commercial real estate
(CRE) markets. ECB Macroprudential Bulletin, N. 19, October 2022.
Horan, A., B. Jarmulska and E. Ryan (2023). Asset prices, collateral and bank lending: the
case of Covid-19 and real estate. ECB Working paper series N. 2823.
Jur?a, P., J. Klacso, E. Tereanu, M. Forletta and M. Gross (2020). The Effectiveness
of Borrower-Based Macroprudential Measures: A Quantitative Analysis for Slovakia.
IMF working paper N. WP/20/134.
Martini, M., P.E. Uriel and A. Jobst (2023). European commercial real estate – selectivity
matters!, Allianz Research Report.
Mian, A. and A. Sufi (2014). House of Debt. University of Chicago Press, Chicago.
Mokas, D. and R. Nijskens (2019). Credit Risk in Commercial Real Estate Bank Loans:
The Role of Idiosyncratic versus Macro-Economic Factors. De Nederlandsche Bank
Ryan, E., A. Horan and B. Jarmulska (2022). Commercial real estate and financial stability
– new insights from the euro area credit register, ECB Macroprudential bulletin, N. 19,
October 2022.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
Annex A: Technical notes
A.1 Breakdowns. The note refers to the following breakdowns:
• Type of property: i) residential; ii) offices and commercial premises; iii) other
commercial (for instance, industrial plants).
• Economic sector of the borrower distinguishing among: i) construction firms;
ii) real estate companies; iii) other firms. The first two categories include entities
whose activities are concentrated in the real estate market and possibly more
exposed to negative developments of the sector. They can be considered as a proxy
of the income-producing RE activities (i.e. the construction/selling of properties
and rental housing) proposed by ESRB Recommendation. The ‘other’ class includes
firms that generally use properties for their own business.7
• Construction investment:8 the breakdown refers to loans granted for the purpose
of investing in construction, including purchase of the land on which the building,
infrastructure and industrial facilities are constructed. In other words, the selection
identifies loans associated to properties ‘under development’, which do not exist at
the time of financing.
• Geographical area where the property is located.
• Lender: we consider both aggregate results for 3 categories of banks.9
A.2 Property valuation in the EU regulation. On the base of the Capital Requirements
Regulation (CRR) art. 208 institutions have to monitor the value of property (eligible for
risk mitigation) on a frequent basis and, at a minimum, once every year for commercial
immovable property and once every three years for residential property. The property
valuation has to be reviewed when available information indicates that its value may
have declined materially relative to general market prices. For loans exceeding 3 million
euros or 5 per cent of the institution funds, the property valuation shall be reviewed at
least every three years. Institutions may use statistical methods to monitor the value
of the immovable property and to identify immovable property that needs revaluation.
Rules on the frequency of valuations and on valuation methodology of collateral
securing NPLs are more stringent factors (par. 7, ECB NPL Guidance 2017); for this
reason, we exclude NPLs in analysis.
A.3 Robustness check on LTV-C. To test the robustness of the LTV-C in AnaCredit,
we compare it with an alternative indicator, where the numerator is the residual
loan amount in AnaCredit and the value in the denominator is the collateral value at
origination adjusted by considering the y-o-y market price changes, distinguishing
This is just a proxy; other kinds of firms in the manufacturing can be involved directly in the real estate sector.
CRE loans with the purpose of ‘Construction investment’ are about 30 per cent in terms of amount and about 40 per
cent in terms of instrument number; half of these are granted to construction and real estate firms, with the vast
majority of the rest going to service company.
The three categories include: 1) Significant banks (excluding ICCREA and CCB groups); 2) BCC (the two banking groups
– ICCREA and CCB – and other cooperative banks not belonging to them); 3) LSI (excluding cooperative banks).
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
among the type of property used.10 We focus on loans granted since 2019, for which we
have all the information at origination (around one third of all loans considered in our
analysis). The results show that the difference in the share of high LTV-C loans between
the two indicators is limited (figure B.2).
since 2019, for which we have all the information at origination (around one third of all loans
considered
in our analysis).
The results
show that the difference in the share of high LTV-C loans
Annex
B: Additional
Figures
and Tables
between the two indicators is limited (figure B.2).
Annex B: Additional Figures and Tables
Figure
valuation
review
breakdowns
Figure
valuation
review
breakdowns
(December 2023 on December 2019; per cent)
a) borrower sector
(December 2023 on December 2019; per cent)
b) construction investment
c) geographical area
Source:
AnaCredit.
Note:
NPL arefrom
excluded
fromEach
thefigure
analysis.
represents
the reviews
distribution
of valuation
Source:
AnaCredit.
Note: NPL
are excluded
the analysis.
represents
thefigure
distribution
of valuation
by the variable
interest
(borrower
type,
construction
investment
and geographical
area). The central
line represents
median value, the
bottom
upper line
reviews
variable
interest
(borrower
type,
construction
investment
geographical
area).
central
lines of the box the 1st and 3rd quartile. The values outside the inter-quantile interval are not shown.
represents the median value, the bottom and upper lines of the box the 1st and 3rd quartile. The values outside the
inter-quantile interval are not shown.
Figure B.2: Loans with high LTV-C – comparison between different current values of collateral
(Per cent; December 2023)
Residential
Commercial – Offices Commercial – Other
AnaCredit
10 Source:
Estimated
AnaCredit. Note: the chart shows the share of high LTV-C loans in the case LTV-C is calculated by using
Given a loan with a collateral h of type k (residential, offices or other commercial) originated in the year 2019, the
AnaCredit data (‘AnaCredit‘) and in the case the indicator is calculated by applying market prices changes to the
adjusted collateral value in 2023 can be estimated as Value_h_23 = V_h_0*(1+change_k_20_19)*(1+change_k_21_20)
collateral value at origination (‘Estimated’). Only loans generated since 2019 are considered in the analysis.
*(1+change_k_22_21)*(1+change_k_23_22), where V_h_0 is the value at the origination reported in AnaCredit and
change_k_x_y is the annual growth rate in the year x for the property type k.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
Source: AnaCredit. Note: NPL are excluded from the analysis. Each figure represents the distribution of valu
reviews by the variable of interest (borrower type, construction investment and geographical area). The centr
represents the median value, the bottom and upper lines of the box the 1st and 3rd quartile. The values outsid
inter-quantile interval are not shown.
Figure B.2: Loans with high LTV-C – comparison between different
Figure B.2: Loans with high LTV-C – comparison
current values between
of collateraldifferent current values of collat
(Per cent; December 2023)
(Per cent; December 2023)
Residential
Commercial – Offices Commercial – Other
AnaCredit
Estimated
Source: AnaCredit. Note: The chart shows the share of high LTV-C loans in the case LTV-C is calculated by using AnaCredit data (‘AnaCredit‘)
and in the case
the indicator
is calculated
by applying
market prices
changesLTV-C
to the collateral
at origination
(‘Estimated’).
Only loans
Source: AnaCredit.
Note:
the chart
shows
the share
of high
loansvalue
in the
case LTV-C
is calculated
generated since 2019 are considered in the analysis.
AnaCredit data (‘AnaCredit‘) and in the case the indicator is calculated by applying market prices changes
collateral value at origination (‘Estimated’). Only loans generated since 2019 are considered in the analysis.
a) Protection value (PV)
a) Protection
(PV) value (PV)
a) value
Protection
Figure
LTVdefinitions
definitions
Figure B.3:
B.3: LTV
(Per cent)
cent)
Figure B.3:
definitions
Figure
B.3: LTV definitions
b) Protection allocated value (PAV)
(per cent) (per cent)
b) Protection
(PAV)value (PAV)
b) allocated
Protectionvalue
allocated
Source:
AnaCredit.Note:
Note:
included
the calculation.
The red
pointthe
indicated
weighted
average
LTV-CLTV
by is
Source: AnaCredit.
areare
included
in theincalculation.
The red point
indicated
weightedthe
average
of LTV-C
by loan of
amount.
amount.
is Note:
calculated
instrument
levels
using
different
definitions
(panels
calculated
atSource:
instrument
levels
using
different
definitions
(panels
LTV-C
consistently
higher
Source:
AnaCredit.
included
calculation.
point
indicated
weighted
average
LTV-C
AnaCredit. Note: NPL are included in the calculation. The red point indicated the weighted average
of LTV-C
the main
impactisisconsistently
on the LTV lower
thanand
60 per
cent;
theimpact
average is
LTV-C
is 55LTV
per lower
cent inthan
December
2023,
in athe
decreasing
trend
formis60
LTV-C
higher
cent;
average
LTV-C
amount.
calculated
instrument
levels
using
different
definitions
(panels
amount.
calculated
instrument
levels
using
different
definitions
(panels
cent, about 12pp higher than the case of PV. In both cases at the numerator we report the loan value, excluding overdrafts and the undrawnthe case
December
2023,
inconsistently
a decreasing
trend
60main
cent,
higher
PV.B.4):
Inthe
cases
thein
LTV-C
higher
and higher
impact
isperforming
theabout
lower
perthan
cent;
average
LTV-C
55ofLTV-C is 5
PAVisthe
LTV-C
impact
LTV lower
60the
cent;
amount,
stated
inconsistently
ESRBisRecommendation.
If we
focus
loans,
thethan
figures
similar
(figure
inaverage
thenumerator
weighted
average
become
centvalue,
theexcluding
share
of loans
greater
80the
is 7.312pp
ofthe
9.8).case
report
undrawn
amount,
asthan
stated
in both
in December
inthe
a decreasing
trend
60overdrafts
cent,60
about
higher
the(instead
cases
per cent
in2023,
December
2023,
inand
a decreasing
trend
cent,
about
higher
of ESRB
In both cas
Recommendation.
wereport
focus only
on performing
loans,
main figures
B.4): in
caseinof
thein the ESR
at the numerator
value,
excluding
overdrafts
the similar
undrawn
asamount,
stated
at the Ifnumerator
we report
the loan
value,the
excluding
overdrafts
and (figure
theamount,
undrawn
asthe
stated
weighted
average
become
54 per
the on
share
of loans
greater
80 per
cent(figure
is 7.3
(instead
ofthe
9.8).
Recommendation.
If we focus
performing
loans,
the main
similar
B.4):cent
in the
case in
the of PAV th
Recommendation.
focus
performing
loans,figures
figures
similar
(figure
B.4):
weighted average
become
54 become
per cent54
share
loans
greater
80 perthan
cent80
is 7.3
(instead
of (instead
9.8).
weighted
average
andofthe
share
of loans
greater
perper
centcent
is 7.3
per cent
of 9.8).
Figure B.4: LTV-C distribution – only performing loans
Figure B.4:
LTV-CB.4:
distribution
– only performing
loans
Figure
LTV-C distribution
– only performing
loans
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025
Source: AnaCredit. Note: NPL are included in the calculation. The red point indicated the weighted average of LTV-C by
loan amount. LTV is calculated at the instrument levels by using two different definitions (panels a and b). In the case of
PAV the LTV-C is consistently higher and the main impact is on the LTV lower than 60 per cent; the average LTV-C is 55
per cent in December 2023, in a decreasing trend form 60 per cent, about 12pp higher than the case of PV. In both cases
at the numerator we report the loan value, excluding overdrafts and the undrawn amount, as stated in the ESRB
Recommendation. If we focus only on performing loans, the main figures are similar (figure B.4): in the case of PAV the
weighted average become 54 per cent and the share of loans greater than 80 per cent is 7.3 per cent (instead of 9.8).
Figure B.4: LTV-C distribution –
only performing loans
Figure B.5: LTV-C greater
than 80 per cent
Figure B.4: LTV-C distribution – only performing loans
(billions of euros; December 2023)
Figure B.5: LTV-C greater than 80 per cent (amount)
Source: AnaCredit. Note. Only performing loans are included. The red point indicated the weighted average of LTV-C
Source: AnaCredit. Note: We distinguish CRE loans in terms of the type of collateral. We consider also a
by loan amount
of different
types of
property
used as collateral.
Source: AnaCredit. Note: Only performing loanscorresponding
are included. to
Thethe caseSource:
AnaCredit.
Note:
We distinguish
CRE loans in terms of the
red point indicated the weighted average of LTV-C by loan amount.
type of collateral. We consider also a mixed category corresponding
to the case of different types of property used as collateral.
Banca d’Italia
Notes on Financial Stability and Supervision No. 44 – January 2025