EY insights on

2020 expected
credit losses
A benchmark across European banks
June 2021
Contents
1 2 3 4 5
Introduction Analysis of the ECL expense Performing Non-performing
full-year 2020 recorded for 2020 exposures: exposures and total
IFRS 9 ECL impacts: coverage ratio coverage ratios
the sample and Stage 2 loan
population proportion

1 2 3 11 17

6 7 8 9 10
Observable trends: Macroeconomic Insights on 2020 How EY teams EY contacts
the product and forecasts ECL disclosures can support you
counterpart lens

18 20 25 31 32

|  EY insights on the 2020 IFRS 9 ECL benchmark  June 2021

Introduction
In early 2021, EY organization performed
a review of the 2020 IFRS 9 expected
credit loss (ECL) disclosures published
by 18 banking institutions headquartered
in Europe.
The purpose of this analysis was to
provide a broad view of how the data
gathered compare across banks,
to present our observations on the
comparisons and to test different ideas
to analyze the data and identify possible
drivers of the trends.
1 |  EY insights on 2020 expected credit losses  June 2021
1
Our focus was: The analysis was presented on an EY IFRS webcast for
clients on 25 March 2021, “IFRS 9 ECL, a benchmark of
• The magnitude of the impact in 2020 (the P/L impact 2020 disclosures and impacts.”
and cost of risk ratios)
A replay of the webcast is available here.
• How the impact has been assessed, with a particular
focus on forward-looking assumptions and overlays For the quarterly analysis that we performed in 2020,
refer to the details below:
• The underlying ECL drivers (focusing on risk
assessment and stage movements) IFRS 9 observations on Q1 impacts and attention points
for half-year reports
• How approaches on staging, scenarios, models and
overlays could be compared IFRS 9: an analysis of Q2 impacts and expectations
Our analysis is based on information collected from
ahead
the IFRS financial statements, earnings presentations IFRS 9 focus areas for year-end 2020
and Pillar 3 regulatory disclosures. A noted limitation
is the significant diversity in terms of content, format
and granularity; accordingly, comparisons between
banks were often challenging. Where this has led to
assumptions, we have referred to these.
A similar analysis was performed by EY each quarter
throughout the year 2020. The insight from the quarter-
on-quarter trends has been incorporated in this analysis.
This document also compares the information disclosed
in the banks’ financial statements, particularly on
how banks explained the impact of COVID-19 in their
ECL amounts.
Analysis of the full-year 2020 IFRS 9
ECL impacts: the sample population 2
The analysis was performed on banking institutions
Figure 1: Geographical location of banks included in sample 1.177b 502b
in France, Germany, Italy, the Netherlands, Spain,
Average total assets Average gross loans to customers
Switzerland and the United Kingdom. These banks are
either classified as a global systemically important
1.000b
institution (G-SII) if headquartered in the European 2.500b
25%
Union, or else classified as a global systemically 800b
2.000b
important bank (G-SIB) if headquartered elsewhere.
1.500b 600b 25%
However, it is important to bear in mind that there are Average

significant differences in total balance sheet size for 1.000b Median

5 400b 25%
the sample population, as well as different business 2 500b 25%
2
mixes in terms of corporate, SME, mortgages and retail
unsecured exposures. These are key considerations for 4 1

both the profit and loss (P/L) impact and the size of the 2 10.1b 3.9b
Average ECL allowance Average ECL P&L charge
impairment allowances. 2

For the sample, the average size of “Gross loans to 24b 12b
customers at amortized cost” is €502b, with four banks 20b
10b
having in excess of €800b and four banks having below 16b
8b
€300b. The EY analysis within this publication focuses
1

12b 6b
on these exposures as the primary scope. 8b
4b
Figure 1 shows the geographical location of banks 4b
2b
included in the sample. 0b

1 All currency conversions to Euro (€) performed as at the exchange rate on

31 December 2020.

2 |  EY insights on 2020 expected credit losses  June 2021

ECL expense recorded for 2020 3
Figure 2: ECL expense for 2020 (€m) 3.1 The overall ECL expense
13,000 The average ECL charge incurred by the banks in 2020
was €3.9b, up from the 2019 average of €2.0b.
11,000 23% All banks emphasized the magnitude of the COVID-19
effect in their communications.
9,000
As shown in Figure 2, the quarter-on-quarter pattern of
13% 20%
recognition of ECL expense varied across the banks, with
9%
7,000 some country trends. For some banks, the ECL charge
in the first half represented as high as 80% or 90% of
25% 14%
5,000 44% 3% 28%
the full-year charge (e.g., those in the UK); whereas for
3% 10% 16%
7% 13% 4% 42%
others, the expense was more evenly spread or even
4%
3,000
27% 34%
10% 25%
22% higher in the second half (e.g., banks in Italy).
33% 8% 21%
56% 15% 8% 16% 31% 15% 14%
63%
16%
15%
32%
17%
19% 17% 10%
12%
25%
19% 14% 9%
In Q1 2020, given the recent pandemic outbreak and
34% 30% 33% 39%
1,000 44% 27% 43% 31%
50%
33%
15%
42%
39%
15% 9% 13% lockdown measures, the effect of COVID 19 was assessed
34% 26% 48% 25% 27%
25% 42%
16% 25% 26%
17% 25% 28% 19%
39%
39% 39% using “top-up” approaches, compared with the normal
“business-as-usual” processes. Significant judgment was
)

)
)

)
)

)
)

)
)

)
)

-1,000
$

$
£

$
€

€
m€

m€

€
€

€
€
(m

(m
(m

(m

(m

(m
(m

(m

involved, and banks’ ECL estimates varied significantly.

(m

(m

(m

(m

(m
(m

(m
(m

1(

2(
B1

B5
B2

B3

B1
B1

B2

B1

B2

B3

B4

B1
B1

B2
B2
B4

nB

nB

For some banks, the overall effect of COVID was not

UK

UK
UK

UK

ss
tch

tch

ch

ch

ch

an
ish

ch

an
ish
UK

l ia

l ia

i
n

rm

rm

Sw
an

an

Du

Du
Ita

Fre
Ita

Fre

Fre

Fre
the largest driver of loss due to significant Stage 3

Ge

Ge
Sp

Sp

Q1 2020 Q2 2020 Q3 2020 Q4 2020 losses reported on single-name defaults in the first half,
attributed to large clients in the Americas and in Asia.

3 |  EY insights on 2020 expected credit losses  June 2021

ECL expense recorded for 2020 (continued) 3
Throughout the year, the approaches were refined
and the dispersion of ECL estimates decreased, but
significant model adjustments and overlays still had to
be used at the year-end to reflect the unprecedented
circumstances (see section 3.4).
The differences in the charges recorded in the second
half of the year were mostly attributable to Q4, when
there was considerable dispersion. Six banks booked a
limited ECL expense, representing 10% or less of their
full-year ECL charge, whereas five banks still booked
a significant charge in Q4 (close or above 30% of their
full-year charge). This second group of banks generally
referred to the deterioration of the environment at the
end of Q4 with new lockdowns and the perspective of
second and third waves. These conditions were harder
than expected at the end of Q3.
Overall, there were very few ECL releases over the
second half, with only four banks in the sample having
booked some releases in Q3 and Q4. Banks either used
overlays to freeze modeled releases or used other forms
of model adjustments aimed at avoiding volatility. This
contrasts with the situation of the four biggest banks in
the US, which reflected a more volatile pattern of ECL
charge over the quarters under the full lifetime ECL
model (95% of the 2020 ECL P/L charge was booked in
the first half of 2020 and three banks showed a release
in Q4 due to macroeconomic improvements).

4 |  EY insights on 2020 expected credit losses  June 2021

ECL expense recorded for 2020 (continued) 3
Figure 3: Cost-of-risk ratio (bps) 3.2 The cost-of-risk (CoR) ratio
The CoR (calculated as the ratio between the ECL charge
250 81 bps 104 bps and the gross loans to customers) doubled in 2020
Average CoR ratio Average CoR ratio
FY2020 First half 2020 (annualized) compared with 2019, with half of the banks concentrated
205 207 200 200 between a ratio of 40 bps and 100 bps (see Figure 5).
200 180 180
160 160 Wide
140 140 dispersion As reflected in Figure 3 and Figure 5, there were some
160
120 120 in first-half significant differences between banks and countries.
154 100 100 2020
150 154
80 80
150
133
138 135 135 60 60
The various drivers for these differences include:

• ► Products, customers and business mix:

40 40
111 20 20
104 105 105 104 0 0
100
84 82
87
81
85
99
90 • Products more sensitive to the economic
77
72 environment (e.g., corporate and commercial
67 64 68 69 64 68
55 53 57 loans), credit cards and other forms of personal
50 44 44
37 39 40 39 42 unsecured lending
33 31
• Exposure to vulnerable sectors
27 26 24 25 26
20 18 19 19 17 18

—
2 • Significant single-name Stage 3 losses in 2020
compared to 2019
B1

B2

B3

B4

B5

B1

B2

B1

B2

B1

B2

B1
nB

hB
nB

hB

hB

hB
UK

ish

tch

an

ss
UK

UK

ish

tch

an
UK

UK

• Geographical footprint: some countries structurally

nc
l ia

nc

nc
l ia

nc

i
rm

rm

Sw
an

an

Du

Du

e
Ita

e
Ita

e
Fr

Fr

Fr

Fr

Ge

Ge
Sp

Sp

have higher loss rates. Banks with exposures in South

FY2019 HY2020 (annualized) FY2020
America and the US were more heavily impacted by
NB. Recalculated ratios which may differ from disclosed CoR the pandemic

• Government support measures

• Macroeconomic projections and assigned weights
• The impact of overlays
Overall, the increase in CoR is relatively balanced
between wholesale and retail portfolios (with retail
catching up in the second half). But corporate and
retail unsecured portfolios attracted the majority of
the increase.

5 |  EY insights on 2020 expected credit losses  June 2021

ECL expense recorded for 2020 (continued) 3
Forward-looking assumptions and overlays represented the strongest drivers as the
Figure 4: CoR ratios — Q4 2020 (annualized) compared with full-year 2019 (in bps)
increase in the CoR was driven by the additional provisions on performing loans,
i.e., Stage 1 and Stage 2 allowances. x4 x2
Italian B1
3.2.1 Various patterns of increase across quarters 180 bps
The spikes observable at half-year in Figure 3 (in dark grey) reflect that more than half
of the banks booked most of the increase in CoR in the first half of the year. For these 160 bps Italian B2
banks, the decrease reflected at year-end (in yellow) does not reflect releases of ECL
allowance in the second half, but rather fewer additions to the ECL allowance compared
with the first half. In contrast, the banks showing a stable CoR between half-year 140 bps
(annualized) and year-end have booked a similar level of provisions throughout Spanish B1

the year. 120 bps

COR Q4 ’20 (annualized)

If we focus on Q4 (annualized), as shown in Figure 4, we can see that some banks Spanish B2
had a very low CoR (lower than for the full year 2019), which reflects that these
100 bps
banks booked the most significant provisions in the first half of 2020. Other banks, in
contrast, exhibited a higher ratio when compared to 2019 (e.g., France reported twice French B2
as much CoR as in 2019 and Italy triple the value of CoR from 2019). 80 bps

These differences in paths throughout the year reveal different situations, different UK B3
German B2
assumptions taken by the banks as well as different levels of model responsiveness 60 bps French B4
and the use of judgmental overlays. These assumptions and overlays were applied to French B3 UK B5
measure the effects of an unprecedented crisis, characterized by a sudden drop in
40 bps German B1
quarterly growth rates as well as a massive level of support measures. Overall, the French B1 UK B1
following need to be considered: Dutch B2
Dutch B1
• The timing and the magnitude of the lockdown measures differed from country to 20 bps UK B4
UK B2
country, with some experiencing stricter measures than others
• Various levels of optimism at half-year, with some banks exhibiting a brighter outlook 0 Swiss B1

for the following periods 0 20 bps 40 bps 60 bps 80 bps 100 bps

• Different overall approaches to forecast and incorporate forward-looking information CoR 2019

into the models, as well as different approaches to overlays NB. Recalculated ratios which may differ from disclosed CoR
• Banks’ own interpretation of guidance issued by several regulators in March 2020, Bubble size reflects Stage 1 + Stage 2 proportion of 2020 P/L charge

recommending banks to take into account the temporary nature of the shock

6 |  EY insights on 2020 expected credit losses  June 2021

ECL expense recorded for 2020 (continued) 3
3.2.2 Country trends Figure 5: CoR ratios (full-year 2020 vs. full-year 2019)
At year-end, most banks were concentrated between a ratio of
40 bps and 100 bps. The dispersion reduced compared with 170 bps x4 x2
Spanish B1
half-year when banks were spread between 30 bp and 207 bp.
160 bps
Figure 5 shows some interesting country trends regarding the level
150 bps
of increase in CoR. Some trends remain consistent between 2019 UK B3
and 2020: Spanish B2
140 bps

• Banks in Spain and Italy stayed within the higher end of the 130 bps
CoR range
120 bps
• The Swiss bank remained at the bottom of the range (driven by 110 bps Italian B2
Italian B1

significant collateralized exposures)

100 bps Dutch B2

COR 2020
For the rest of the countries, the COVID-19 crisis has increased the
UK B1
dispersion between countries and sometimes between banks: 90 bps UK B4

• UK banks experienced a more significant increase compared 80 bps

UK B2
UK B5

with other European banks (CoR has tripled) but most of them 70 bps French B2
French B4
converge towards an average of 83 bp at year-end
60 bps German B2
• French, German and Dutch banks are more dispersed (between 50 bps Dutch B1
35 bps and 90 bps), partly driven by some significant Stage 3 German B1
French B3
losses at some banks in Q1 and Q3. And the average increase is 40 bps
lower, around two times 30 bps French B1
Swiss B1
In comparison, the CoRs of the four biggest US banks tripled on 20 bps
average (178 bps on average), after a significant spike at half-year
10 bps
(when the annualized CoR was a multiple of five).
0 20 bps 40 bps 60 bps 80 bps 100 bps

COR 2019

NB. Recalculated ratios which may differ from disclosed CoR

Bubble size reflects Stage 1 + Stage 2 proportion of 2020 P/L charge

7 |  EY insights on 2020 expected credit losses  June 2021

ECL expense recorded for 2020 (continued) 3
3.3 An increase in ECL driven by The increase in Stage 1 and Stage 2 provisions was In Q1, banks generally used overlays, as no revised
driven by forward-looking assumptions, i.e., the revision scenario was publicly available in time for the Q1 closing
Stage 1 and Stage 2 provisioning
of the macroeconomic scenarios, in particular, the (in particular from regulators) and uncertainty was
As noted in Figure 6, Stage 1 and Stage 2 provisioning baseline scenario (see section 7 on page 20). extreme. Banks used revised scenarios in Q2, which
represented 43% of the 2020 ECL P/L charge for 2020, generally resulted in further increases in Stage 1 and
compared with only 10% in 2019. 2 provisioning. There were limited updates required in
Q3, as the banks in this quarter experienced better than
expected performance, but more uncertainties arose at
Figure 6: Stage 1 and Stage 2 proportion of total ECL P/L charge (in %)
the same time.

90%
As mentioned in section 3.2.1, Q4 was mixed and
resulted for some banks in further deterioration of the
81%
assumptions due to the new imposed national measures,
70%
the threat of second and third COVID-19 waves and
longer vaccination periods. This resulted in three main
52%
50%
50% 52%
48% 45%
trends in Q4:
45%
41%
39%
35%
38% 38%
1. French and Italian banks, together with one German
32%
30% 29% bank, substantially increased their Stage 1 + Stage 2
25%
23%
ECL allowances (40% of the increase for the full year)
14%
11%
10% 2. A few banks released Stage 1 and Stage 2 ECL
allowances due to scenario improvements (between
20% and 50% of the increase accumulated in prior
-10%
quarters)
3. The rest of the banks exhibited little movement, with
-30%
modeled releases neutralized by overlays.
B1

B2

B3

B4

B5

B1

B2

B1

B2

B1

B2

B1
nB

hB
nB

hB

hB

hB

Generally, the banks showing the lowest proportions

UK

ish

tch

an

ss
UK

ish

tch

an
UK
UK

UK

nc
l ia

nc

nc
l ia

nc

i
rm

rm

Sw
an

an

Du

Du

e
Ita

e
Ita

of Stage 1 and Stage 2 ECL expense at year-end either

Fr

Fr

Fr

Fr

Ge

Ge
Sp

Sp

FY2019 HY2020 FY2020 experienced significant single-name Stage 3 losses

during the year, or they have already released a portion
of their Stage 1 and 2 allowance in Q3 or Q4.

8 |  EY insights on 2020 expected credit losses  June 2021

ECL expense recorded for 2020 (continued) 3
3.4 The use of management overlays risk counterparties). As such, banks booked very 3. Sector idiosyncrasies were amplified by the crisis
significant adjustments in order to model better the and adjustments were applied in order to have an
A significant component of the Stage 1 and 2 increased
losses expected in this crisis, including the effect of appropriate differentiation in the severity of projected
provisioning resides in overlays. They proved to be
government support programs. default rate conditions for different industry sectors.
important factors as early as Q1 and evolved in nature
over the quarters. 2. The historical correlation between GDP and other key It is not yet clear how or when these overlays will be
economic variables and future losses was built without released, but generally, banks confirmed that the
Some significant overlays booked in Q1 were
considering the effect of governmental support releases would be limited before the support measures
subsequently “recycled” in the revision of
measures. In practice, this was adjusted by averaging have come to an end.
macroeconomic scenarios. But overlays evolved into a
inputs or correcting the outputs.
mix of post-model adjustments and uncertainty overlays
and remained substantial. Figure 7: Impact of overlays on 2020 CoR
As there is no taxonomy available for overlays, diversity 140 bps
exists with regards to what banks assign against this 130 bps
term. However, based on what 11 banks described as 120 bps
overlays at year-end, we observed (Figure 7) that the
110 bps
CoR attributable to overlays represented more than 85%
100 bps
on average of the CoR attributable to Stage 1 and Stage
90 bps
2 (the lowest level being around 20%).
80 bps
Overlays resulted in a net increase in the ECL balance,
70 bps
but with significant offsetting effects between negative
60 bps
post-model adjustments and positive uncertainty
50 bps
overlays. Significant additional overlays were booked at 60%
year-end to avoid model releases and to take into account 40 bps
111%
97%
lag effects in expected defaults as well as significant 30 bps
33% 91%
uncertainties. 20 bps 76%
30% 85% 110%
10 bps 19% 24%
There were three main reasons for overlays:
0
1. Models were operating outside the boundaries
B1

B2

B3

B4

B5

B1

B1

B2

B2
hB

hB
of data used to calibrate them. Several banks
UK

ish

tch
UK

UK

tch

an
UK

UK

nc

nc

rm
an

Du

Du

e
Fr

Fr

Ge
referred to models providing unrealistically high

Sp
default rates (most often for wholesale and low- CoR (total) S1 + S2 share in CoR Overlays share in CoR (as disclosed) % Overlays as a proportion of S1 + S2 CoR

9 |  EY insights on 2020 expected credit losses  June 2021

ECL expense recorded for 2020 (continued) 3
3.5 What is the CoR outlook for 2021?
All banks have projected a 2021 CoR outlook below that
for 2020, with all of them stressing the high level of
uncertainty.
The different trends observed in the 2021 CoR guidance
published by banks were:

• A normalization of the ECL charge and a return closer

to pre-pandemic levels

• A return to ‘through-the-cycle’ levels of CoR

• A level of impairment charge in 2021 materially below
that of 2020
or

• A level of impairment charge below 2020 but

still elevated

10 |  EY insights on 2020 expected credit losses  June 2021

Performing exposures: coverage ratio
and Stage 2 loan proportion 4
Figure 8: Stage 1 + Stage 2 coverage ratio 4.1 Analysis of Stage 1 and Stage 2 coverage ratios
Stage 1 + Stage 2 ECL allowance/Stage 1 + Stage 2 gross loans to customers The Stage 1 and Stage 2 (performing) coverage ratio (Stage 1 and Stage 2 ECL
x4 x2
Spanish B2 allowance divided by Stage 1 and Stage 2 gross loans to customers) experienced an
1.4% UK B3
increase of 1.5 times on average over the year, with that of a few banks tripling or more
1.3% (see Figure 8).

1.2%
This represented a very significant increase given the size of the underlying exposures
Spanish B1 (Stage 1 makes up 85%–90% of the total loan book).
1.1%
UK B4 An interesting measure is to compare the Stage 1 and Stage 2 ECL allowance to one
Italian B1
1.0% year of pre-pandemic Stage 3 losses (refer to Figure 10). This ratio indicates how many
years of Stage 3 losses the Stage 1 and 2 allowances cover. At year-end 2020, using
0.9% the 2019 Stage 3 losses as a reference, the Stage 1 and 2 ECL allowance represented
2.4 times the Stage 3 losses, compared with 1.4 at year-end 2019. However this
0.8%
FY 2020

average increase hides a wide dispersion between banks, with ratios varying from one
French B4 French B1
0.7% UK B1 to six.
French B2 The distribution of coverage ratios was already quite wide in 2019. This is particularly
0.6%
French B3
Italian B2
visible in the UK and Germany.
0.5%
UK B5 The dispersion is driven by the characteristics of products (loan to value, collaterals,
UK B2
0.4% guarantees and origination criteria) and local economics, but the increase may also
Dutch B1
German B2 reflect differences in IFRS 9 approaches.
0.3%
German B1 Banks with lower coverage ratios at the beginning of the year have often increased
Dutch B2
0.2% the most their CoR in the course of 2020. In a few cases, banks have caught up with
their counterparts. This is reflected on Figure 9 which compares the coverage ratios on
0.1%
Swiss B1 performing loans to customers at year-end 2019 (on the horizontal axis) and by how
much banks have multiplied their 2019 CoR in 2020 (on the vertical axis).
0.1% 0.2% 0.3% 0.4% 0.5% 0.6% 0.7% 0.8% 0.9% 1.0% 1.1% 1.2%
FY 2019
Bubble size reflects Stage 2 proportion of Stage 1 + Stage 2 loans at YE 2020

11 |  EY insights on 2020 expected credit losses  June 2021

Performing exposures: coverage ratio and
Stage 2 loan proportion (continued) 4
Figure 9: Increase in CoR (2020 CoR divided by 2019 CoR) compared to year-end 2019 Stage 1 + Stage 2 Figure 10: How many years of 2019 S3 losses does the Stage 1 + Stage 2 allowance represent?
coverage ratio (Stage 1 + Stage 2) ECL allowance/2019 S3 losses

6.1
6.0
4.4 UK B4
5.5
4.2
5.0
4.0
4.5
Dutch B2
3.8
4.0

3.6 3.5 3.6

3.5
3.2 3.1
3.4 3.0
3.0
2.9
UK B2
Increase in CoR in 2020
2020 CoR/2019 CoR

3.2 2.5 2.3

UK B1 2.2
2.1 2.1
2.0
3.0 2.0
1.8
1.9
1.7
1.7 1.7 1.7
1.4 1.5 1.5 1.5
2.8 French B4
1.5 1.4
1.2 1.3 1.3
UK B3 1.2 1.1 1.1
German B2 1.0 1.0
UK B5 1.0
2.6 Dutch B1 0.7 0.8
0.7
0.5
2.4
German B1 0.0
2.2
French B3

B1

B2

B3

B4

B5

B1

B2

B1

B2

B1

B2

B1
French B1

nB

hB
nB

hB

hB

hB
UK

ish

tch

an

ss
UK

tch

an
UK

ish
UK

UK
2.0

nc
l ia

nc

nc
l ia

nc

i
rm

rm

Sw
an

an

Du

Du

e
Ita

e
Ita

e
Fr

Fr

Fr

Fr

Ge

Ge
Sp

Sp
1.8 2019 2020
Italian B2 French B2 Spanish B2
1.6

1.4 Spanish B1
Italian B1
1.2

0.1% 0.2% 0.3% 0.4% 0.5% 0.6% 0.7% 0.8% 0.9% 1.0% 1.1% 1.2%
YE 2019 �S1 + S2 coverage ratio

Bubble size reflects Stage 2 proportion of Stage 1 + Stage 2 loans at YE 2019

12 |  EY insights on 2020 expected credit losses  June 2021

Performing exposures: coverage ratio and
Stage 2 loan proportion (continued) 4
Figure 11: Stage 2 gross loans to customers as a proportion of Stage 1 and Stage 2 loans to customers 4.2 Stage 2 loan proportion
The increase in Stage 1 and Stage 2 coverage ratios was
30% driven in large part by an increase in Stage 2 exposures.
The average proportion of Stage 2 exposures has
25% increased from 7.6% to 11% of performing exposures. But
22% there is considerable variety in trends throughout 2020,
20% 19%
as well as some quite different starting points before
the crisis, as shown in Figure 11.
16%
16%
15%
15%
A first group of banks has recognized a sharp increase
12%
in Stage 2 exposures in the first half (in dark grey), with
11%
11%
10%
10% 10% some of them seeing a decrease in the proportion in the
10%
8%
7% 7%
9% 8% second half. A couple of banks have experienced a larger
5% 5%
increase in the second half whereas a few others show
5% limited increase compared with 2019.
This reflects local differences in national economic
0% outlook due to local differences in the spread of the virus,
containment measures and government support. But it

B1

B2

B1
B1

B2

B3

B4

B5

B1

B2

B1

B2

1
nB

hB
nB

hB

hB

hB

an

ss
UK

ish

h
UK

ish

an
UK
UK

UK

tc

nc
l ia

tc

nc

nc
l ia

nc

i
may also reflect that banks have designed their Stage

rm

rm

Sw
an

an

Du

Du

e
Ita

e
Ita

e
Fr

Fr

Fr

Fr

Ge

Ge
Sp

Sp

2 triggers with varying levels of sensitivity to forward-

FY2019 HY2020 FY2020
looking information and may have used different ways
to tackle the lack of visibility due to moratoria and other
support measures, through the use of overlays and
portfolio approaches.
Based on regulators’ published guidance, large-
scale moratoria were generally not considered to be
forbearance measures and did not trigger Stage 2
transfers on a standalone basis. Then, because moratoria
and guaranteed loans improved the apparent credit
quality of borrowers (no arrears and/or larger cash
balances), the effects of the crisis were not observable in

13 |  EY insights on 2020 expected credit losses  June 2021

Performing exposures: coverage ratio and
Stage 2 loan proportion (continued) 4
the customers’ financial performance. Hence, it was very Figure 12: Loans and advances subject to legislative and non-legislative moratoria (unexpired)
challenging for banks to identify a significant increase in
100 15%
credit risk at an individual-borrower level.
13%
90
Movements in forward-looking PD were, therefore, the 11%
10%
11%
10%
main driver for transfers as opposed to the more lagging 80
7% 7% 7% 7%
backstops based on delinquency and forbearance. Banks 6% 6%
5% 5%
6%
5%
70
also supplemented their quantitative and qualitative 4%
3% 3%
2% 3% 2%
2% 2% 1%
criteria with targeted analyses for vulnerable sectors 60 1% 1% 1% 1% 1%
0% 0% 0% 0% 0%
and for borrowers that had weaker pre-crisis ratings.
50
They also strengthened the level of client monitoring
-5%
implemented by credit managers and refined the analysis 40

of accounts movements to differentiate cash inflows

30
stemming from support measures versus actual business -10%

activity. When modeled PDs were not increasing, banks 20

used overlays to increase the ECL, but this was not -15%
10
always visible at the level of Stage 2 exposures as the
overlays were additions to the ECL allowance and actual — -20%
UK B1, HY 20
YE 20

UK B2, HY 20
YE 20

UK B3, HY 20
YE 20

UK B4, HY 20
YE 20

UK B5, HY 20
YE 20

Spanish B1, HY 20
YE 20

Spanish B2, HY 20
YE 20

Italian B1, HY 20
YE 20

Italian B2, HY 20
YE 20

Dutch B1, HY 20
YE 20

Dutch B2, HY 20
YE 20

French B1, 20'Q2

YE 20

French B2, HY 20
YE 20

French B3, HY 20
YE 20

French B4, HY 20
YE 20

German B1, HY 20
YE 20

German B2, HY 20
YE 20
exposures were not transferred.
As shown in Figure 12, most banks highlighted a material
decrease in large-scale moratoria at year-end and the
overall good performance of customers for whom the
moratoria had expired. But they also stressed that Households Non-financial corporations (non-SME) SME Other % of total gross loans subject to moratorium
significant uncertainties remained over the length and
scale of support measures.

14 |  EY insights on 2020 expected credit losses  June 2021

Performing exposures: coverage ratio and
Stage 2 loan proportion (continued) 4
Figure 13: Newly originated loans and advances provided under newly applicable public guarantee schemes
As shown in Figure 13, originated loans under newly
applicable public guarantee schemes represented
45 30% significant volumes in some countries at year-end (e.g.,
24%
France). The graph shows the diversity in the application
40
of this measure between countries: although some

% of exposure not covered by public guarantee

20%
35
18%
16%
18% countries show a limited impact on banks’ balance sheets
14%
13.17% 13%
12% 13% (such as in the Netherlands or Germany), other countries
11%
30
8%
10% 10% (like France) have widely used this measure to support
Currency unit, bn

7%

25 4% 4%
the economy through banks’ balance sheets. The level
0%
of guarantee provided by the government also differs
20 between countries and between banks as some schemes
involved a different level of guarantee depending on the
15
-10% size of the company. Guaranteed loans involve similar
10 uncertainties as moratoria, with longer-term effects,
-20%
as they also often involve payment holidays and have
5 increased the level of leverage.
—
-30%
UK B1, HY 20
YE 20

UK B2, HY 20
YE 20

UK B3, HY 20
YE 20

UK B4, HY 20
YE 20

UK B5, HY 20
YE 20

Spanish B1, HY 20
YE 20

Spanish B2, HY 20
YE 20

Italian B1, HY 20
YE 20

Italian B2, HY 20
YE 20

Dutch B1, HY 20
YE 20

Dutch B2, HY 20
YE 20

French B1, 20'Q2

YE 20

French B2, HY 20
YE 20

French B3, HY 20
YE 20

French B4, HY 20
YE 20

German B1, HY 20
YE 20
Households Non-financial corporations (non-SME) SME Other % of exposure not covered by public guarantee

NB: for Italian B2 and French B3, the guaranteed amounts were not disclosed at FY 2020.
For Italian B1, Italian B2, French B1, French B3 and French B4, the counterparty allocation was not available at the date of analysis. Therefore, the full balance has been
allocated to “other”.

15 |  EY insights on 2020 expected credit losses  June 2021

Performing exposures: coverage ratio and
Stage 2 loan proportion (continued) 4
Figure 14: Increase in Stage 2 loans to customers compared with increase in CoR (2020 CoR divided by 4.3 Increase in Stage 2 loan proportion compared with increase
2019 CoR)
in CoR
We also considered how the increase in Stage 2 exposures relates to the level of
increase in CoR (similar to the analysis published in the ECB supervision blog “Who pays
2.6 UK B4 the piper calls the tune” by Elizabeth McCaul, 4 December 2020).
Figure 14 shows that some banks have experienced a significant increase in CoR
2.4 (reflected as a multiple of 2019 CoR), which does not necessarily reflect in the increase
in Stage 2 transfers. This is due to the challenges raised by the estimate of the impact
Increase in Stage 2 proportion of S1 + S2 loans

of support measures on borrowers’ viability and the use of material overlays (not always
2.2
translated into Stage 2 transfers).
Italian B1
UK B1 Some catch-up increase in Stage 2 exposures may be observed in 2021 as banks
2.0
identify more significant deterioration and apply more transfers in line with the amount
French B4
of ECL increased through overlays in 2020.
1.8 UK B2
Dutch B1

1.6
German B1 Dutch B2

1.4 Spanish B1 UK B3

French B2
German B2
1.2
Spanish B2 French B1

UK B5
1.0
French B3

1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Increase in CoR (CoR2020/CoR2019)

Bubble size reflects Stage 2 proportion of Stage 1 + Stage 2 loans at YE 2020

16 |  EY insights on 2020 expected credit losses  June 2021

Non-performing exposures and
total coverage ratios 5
Figure 15: Total coverage ratio (all stages) 5.1 Stage 3 transfers and write-offs
Total ECL allowance/Total gross loans to customers The challenges faced by banks to look through the effect of support measures to assess
the resilience of borrowers is also reflected in the relative stability of Stage 3 transfers
x4 x2 Italian B1
3.8% US B4 and write-offs.
Spanish B2
3.6% Net transfers to Stage 3 increased slightly, but this was generally driven by significant
3.4% single-name losses not directly linked to the COVID-19 crisis. Write-offs were generally
3.2% stable in 2020 compared with 2019 (including for US banks).
3.0% Italian B2
5.2 Total coverage ratios (Stage 1, Stage 2 and Stage 3)
US B1
2.8% Spanish B1

2.6% French B4 French B2 As shown in Figure 15, the increase in total coverage is more limited than for
2.4% performing loans only. The magnitude of the increase of Stage 1 and 2 provisions is
US B3
UK B3
UK B5
diluted by the impact of Stage 3 for which the coverage ratio was stable or decreased.
2.4%
FY 2020

2.0% US B2 In contrast, US banks doubled their total coverage ratios (with a wide range between
French B1
1.8% UK B4 1.3% and 2.7%). This is partly due to quicker write-offs and a lower proportion of
1.6%
non-performing loans (0.85% on average), meaning that credit-impaired exposures are
Dutch B2 French B3
UK B1
normally better-quality and attract a lower ECL allowance.
1.4% UK B2

1.2% The level of non-performing loans will be a key indicator of how economies perform
German B2
1.0% when support measures come to an end in the coming months, and it is therefore
0.8%
German B1 relevant to observe how banks compare at year-end. This is reflected in Figure 15 by
Dutch B1
the “bubble” size, with Italy and Spain showing the highest levels (respectively 4.4% and
0.6%
3.9%), followed by France and Germany (3.1%), the Netherlands (2.8%), the UK (2.6%)
0.4% Swiss B1
and Switzerland (0.8%).
0.2%

0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% 4.5%
FY 2019

Bubble: Bubble size reflects Stage 3 proportion of gross loans to customers at YE 2020

17 |  EY insights on 2020 expected credit losses  June 2021

Observable trends: the product
and counterpart lens 6
Figure 16: Product segmentation: Stage 2 loans as a proportion of Stage 1 and Stage 2 loans 6.1 Product segmentation
Only a small sample of banks provided the ECL impact
Wholesale (excluding SMEs) Household
split between products. This means that determining
30%
25%
whether the drivers are wholesale exposures, smaller
25% 24% ∆ +91% ∆ +24% businesses, retail unsecured or mortgages is not
20%
21% 20%
17.9% possible for most reporters. In our analysis of the
16% 16% 16%
15% 14.5% 14.6% product segmentation, we obtained the segment data
15% 14% 13% 14% 13.3%
13%
11% 11%
12% from either the annual reports or Pillar III disclosures.
10.5% 10.5% 10.4%
10% 8%
9% 10% 9.0% 8.4% Personal unsecured loans mainly consist of credit cards
6.6% 6.8%
5% 5% 5%
5%
5.5% 5.0% 4.9% 4.3% 4.9% and overdrafts.
2.0%
0%
0%
For banks that provided detailed information, as it can
UK B2

UK B3

UK B4

UK B5

Spanish B1

Spanish B2

Italian B1

Italian B2

Dutch B1

Dutch B2

French B1

French B2

French B4

German B1

German B2

Swiss B1

UK B1
be seen in Figure 16, wholesale portfolios showed the
UK B2

Italian B1

Italian B2

Dutch B1

Dutch B2

French B1

French B2

French B3

French B4

German B1

German B2
UK B3

UK B4

UK B5

Spanish B1

Spanish B2
largest increase in Stage 2 (+88%), with SMEs reflecting
FY2019 FY2020
a consistent but more nuanced trend (+48%). The
FY2019 FY2020
more nuanced increase for SMEs, along with personal
SMEs Personal unsecured
unsecured exposures, reflect the difficulty banks
experienced in identifying early signs of deterioration for
45% individuals and smaller businesses.
40.7% 45%
40%
36.0% ∆ +48% 40%
36.3%
∆ +34% Wholesale portfolios also exhibited the highest increase
35%
35%
30%
26.8%
in the Stage 1 and Stage 2 (performing) coverage ratio
25.1% 30%
25% 22.2% 24.6% 25% 23.4% 23.8% with an average increase of 116%, followed by personal
22.0%
20%
14.9% 15.9% 20%
20.0% unsecured loans (average increase of 68%).
15% 13.1% 13.0% 14.1%
10.4% 10.9% 11.7% 11.0% 10.8% 15%
10% 9.1%
10%
9.9% The increase in total coverage ratios was consistent
5%
5% 4.5% but less pronounced when compared to the
0%
0% Stage 1 + Stage 2 ratio, with a mild increase for
UK B2

UK B3

UK B4

UK B5

Spanish B1

Spanish B2

Italian B1

Italian B2

Dutch B1

Dutch B2

French B1

French B2

French B3

French B4

German B1

German B2

Swiss B1

UK B1

UK B2

UK B3

UK B4

UK B5

Dutch B2

Swiss B1
household exposures.

FY2019 FY2020
FY2019 FY2020

18 |  EY insights on 2020 expected credit losses  June 2021

Observable trends: the product and
counterpart lens (continued) 6
6.2 Vulnerable sectors Almost all banks added an emphasis on vulnerable sectors in their disclosures. From
our analysis, the top five sectors (in terms of vulnerability) presented in the disclosures
were: non-food retail, aviation, oil and gas, transportation and leisure/hospitality (see
Figure 17: Number of banks citing specific vulnerable exposures Figure 17). From these sectors, oil and gas, hospitality and retail represented the
largest exposures. Across banks, as shown in Figure 18, the proportion of vulnerable
Agriculture and food sectors’ exposure to the total wholesale book varied from 7% to 16%. However, the
FY 2020
Transportation and 16 Automobiles and information provided is not consistent across the population. For example, some banks
logistics (excl. aviation) components
14 disclosed the gross carrying amounts of their vulnerable sectors’ exposures whereas
12 others have considered their exposure at default. In other cases it was not immediately
10 clear what type of exposure was presented.
Textiles Aviation
8
Figure
Figure 18: Gross
18: Gross exposure
exposure to topto top
five five vulnerable
vulnerable sectors as %sectors as % ofloans
of total wholesale total wholesale loans
6
4 18.0% 27b
29b

Sports,
Education 2 Commodity traders 16.0% 13b
2.5%
and Media 0 19b
2.9% 97b
14.0%
1.7%
1.2%
62b 1.5% 4.6% 50b
12.0% 2.6%
0.6%
Construction, real 10.0%
8b 3.4% 5b
1.8% 24b
Retail estate development 3.7%
5.5%
11b
18b 1.6%

(net)
(non-food) and infrastructure 8.0% 1.4% 3.2% 4.7%
3.2%
4.3% 2.2% 4.7%
2.5% 3.3%
6.0% 2.5%
1.1% 2.1%
4.9% 9.0%
4.0% 0.6%
5.8%
Oil and gas Health 3.7% 6.6%
2.3% 3.0%
5.8%
4.7% 4.1% 3.0%
2.0% 3.9%
2.1% 1.8% 1.7%
1.2% 1.0% 0.8%
Metals and mining Hospitality, leisure and tourism 0.0% 0.4%
UK B1 UK B2 (€) UK B3 (€) UK B4 (€) UK B5 ($) ES 2 (€) NL B1 (€) NL B2 (€) FR B2 (€) FR B3 (€) FR B4 (€) CH B1 ($)

Oil & Gas Hospitality, leisure and tourism Retail (non-food) Aviation Transportation & logistics
* This graph is based on the sectorial disclosures provided by 15 banks in our sample
= Total gross exposure to top 5 vulnerable = Total On- and Off- balance = Total exposure
sectors in local currency billions* sheet gross exposure (not further defined)

* Note that data is not fully comparable, as the basis of preparation differs (e.g., gross vs. net book values or Exposures At
Default) and, in some instances, is not disclosed
** This graph is based on the quantitative sectorial disclosures provided by 12 banks in our sample

19 |  EY insights on 2020 expected credit losses  June 2021

Macroeconomic forecasts 7
Figure 19: Year-end (left) vs. half-year (right) 2020 GDP forecast in the UK (2019=100) 7.1 Forecasts for 2021 and beyond
105 105 The GDP forecasts show an improved outlook for
UK B1 BoE (Feb 2021) UK B1 BoE (Aug 2020)
UK B2 UK B2 BoE (May 2020) 2021 and beyond compared with 2020, as shown in
UK B3 UK B3 EC (Jul 2020)
UK B4 UK B4
Figures 19 and 20. The trend is similar across both the
100 100
Eurozone and the UK, with a much quicker recovery
predicted for 2022 in the UK.
95 95 As we have mentioned in section 3, the increase in Stage
1 and Stage 2 provision is driven by forward-looking
information. The revision of the baseline scenario has
90 90
been the primary driver as evidenced by the revised
pattern of growth rates that banks incorporated into their
85 85 models. This was one of the most challenging areas in
2019 2020 2021 2022 2023 2019 2020 2021 2022
the first half of the year, especially in Q1 as no revised
Source: Bank of England (BoE), Monetary Policy Report, February 2021 Source: European Commission (EC), Summer 2020 Economic Forecast, July 2020; Bank of scenario was publicly available in time for the Q1 closing
England (BoE), Monetary Policy Report, August 2020; Bank of England, Monetary Policy
Report, May 2020 (in particular from regulators) and uncertainty was
extreme. In Q1, banks generally used overlays and only
Figure 20: Year-end (left) vs. half-year (right) 2020 GDP forecast in the Eurozone (2019=100) revised their scenarios in Q2.
110 French B1 BoE (Feb 2021) 110 French B1 ECB mild (Sep 2020) There was little update required for Q3 as economies
French B2 ECB mild (Mar 2021) French B3 ECB baseline (Sep 2020)
French B4 ECB baseline (Mar 2021) French B4 ECB severe (Sep 2020)
generally performed better in the third quarter than
105
German B1 ECB severe (Mar 2021) ECB mild
105
German B1 ECB baseline (Jun 2020)
ECB mild
expected at half-year, but at the same time, more
Spanish B1 Italian B1
Swiss B1 generally uncertainties arose at the end of the quarter.
100
ECB severe
100 Q4 generally resulted in further deterioration of
the assumptions due to new lockdowns and the
95 95 perspective of harder second and third waves and longer
ECB severe
vaccination processes.
90 90
Based on the banks’ forecasts, the 2019 level is reached
again at the end of 2022 or in 2023. Also, as can be
85
2019 2020 2021 2022 2023
85
2019 2020 2021 2022 seen in Figures 19 and 20, there is an overall good
convergence, but with banks having slightly lower
Source: European Central Bank (ECB), Eurosystem Staff Macroeconomic Projections, Source: European Central Bank, Eurosystem Staff Macroeconomic Projections,
March 2021; Bank of England (BoE), Monetary Policy Report, February 2021 September 2020; European Central Bank, Eurosystem Staff Macroeconomic Projections,
June 2020
projections than regulators’ revised forecasts.

20 |  EY insights on 2020 expected credit losses  June 2021

Macroeconomic forecasts (continued) 7
7.2 Scenarios Figure 21: CoR of alternative scenarios: actual/baseline/mild downside/severe downside
The uncertain environment has also given prominence to 260 bps +52% (**)

the use of multi-scenarios in ECL estimates. 240 bps -52% (*)

+65%
Half of the banks in the sample provide sensitivity analysis 220 bps
showing what the ECL allowance would amount to under 200 bps
-7%

each individual scenario if weighted at 100%. Although +18%

180 bps
the number of banks is more limited, our analysis shows
+82% +91%
significant differences between the reported ECL and what 160 bps
+68%
it would be under the alternative downside scenarios, 140 bps -14% -9%
as shown in Figure 21. The graph compares the CoR as 120 bps +43%
-14%
reported, and then how it would look like if each alternative 100 bps
-15% -1%
+1%
scenario was weighted at 100% (we have focused on
80 bps
baseline and downside scenarios only). We observed that +41%
-7%
the difference between the effect of the central scenario 60 bps +28%

in isolation and the reported probability weighted ECL is 40 bps +52%

-28%
around 15%. Meanwhile, the difference between the effect 20 bps
of the severe downside scenario and the reported ECL is 0 bps
50% on average. UK B1

UK B2

UK B3

UK B4

UK B5

Dutch B1

Dutch B2

German B1

Swiss B1

US B1

US B3
There are very significant differences that are difficult to
analyze as they may be due to a combination of factors,
such as: Actual (probability weighted) Mild downside scenario (weighted at 100%) (*) in the analyst’s call

• Assumptions: consideration needs to be given to the Severe downside scenario (weighted at 100%) Baseline scenario (weighted at 100%) (**) mild downside minus baseline

level of stress applied in the downside scenarios, the

dispersion of alternative scenarios used, but also the
differences in assumptions taken in the central scenario.
• Weights: As reflected in Figure 22, scenarios are
balanced by different weights. An extreme scenario with
an assigned weight of 5% may be less impactful than a
milder one with a higher weight.
• Products and geographies: the range of products vary
as well as the trends and assumptions observed across
different geographies.
Overall, it was noted that different approaches were used
across the analyzed population.
This also applied to the US where, for example, one
bank showed an increase on the mild downside over the
reported ECL allowance close to 50% whereas another
bank projected an increase on the severe scenario
compared with the reported ECL close to 20%.
The use of dissimilar sensitivity analysis made
comparisons challenging: some banks compare the ECL
allowance under each scenario weighted at 100% with
the reported ECL allowance, whereas others use the
baseline ECL allowance as a benchmark; also, although
some banks recalculated the full ECL allowance under
each scenario, other banks recalculated it for Stage 1 and
Stage 2 only.

21 |  EY insights on 2020 expected credit losses  June 2021

Macroeconomic forecasts (continued) 7
7.3 MES weightings
Scenarios are then balanced by different weights.
Figure 22 shows how banks have rebalanced the
weights of alternative scenarios to reflect increased
uncertainties. We observed that half of the banks had
decreased the weight of the baseline to put more weight
on the downside (including with the addition of a new
alternative downside scenario).
There is a variety of approaches in the scenario
weightings:

• Some banks have retained the same weightings,

considering that they remained appropriate given the
revision of scenarios.

• One bank explained the weights are designed to react

to the point in the cycle, which means that the lower
case gets less weight in the current circumstances.

• Others had a more ad hoc judgmental approach, i.e.,

some reduced the number of scenarios and removed
the most extreme ones, but others added additional
downsides.

22 |  EY insights on 2020 expected credit losses  June 2021

Macroeconomic forecasts (continued) 7
Figure 22: Macroeconomic scenario (MES) weightings: Q4 2019, Q2 2020 and Q4 2020

UK B1 MES weightings UK B2 MES weightings UK B3 MES weightings

80% 80% 80%

NB. UK B2
retained the same
60% weighting for all 60%
60%
three reported
quarters
40% 40%
40%

20% 20%
20%

0% 0%
0% Upper Base Lower Lower Upper Upper Base Lower Lower
Upper Base Lower Lower case 1 case case 1 case 2 case 2 case 1 case case 1 case 2
case 1 case case 1 case 2
20 Q4 19 Q4 20 Q2 20 Q4
19 Q4 20 Q2 20 Q4

UK B4 MES weightings Spanish B1 MES weightings Spanish B2 MES weightings

80% 80% 35%

NB. This NB. Spanish B2
represents the retained the same
60% 60% 34%
UK MES weighting weighting for
used by Spanish 19’Q4, 20’Q2 and
40% 40% B1 34% 20’Q4

20% 20% 33%

0% 0% 33%
Upper Upper Base Lower Lower Upper Upper Base Lower Lower Lower
Upper Base Lower
case 2 case 1 case case 1 case 2 case 2 case 1 case case 1 case 2 case 3
case 1 case case 1
19 Q4 20 Q2 20 Q4
19 Q4 20 Q2 20 Q4
20 Q4

23 |  EY insights on 2020 expected credit losses  June 2021

Macroeconomic forecasts (continued) 7
Figure 22: Macroeconomic scenario (MES) weightings: Q4 2019, Q2 2020 and Q4 2020 (continued)

Dutch B1 MES weightings Dutch B2 MES weightings Swiss B1 MES weightings

80% 80% 80%
NB. Dutch B1 NB. Dutch B2
retained the same retained the same
60% weighting for 60% weighting for 60%
19’Q4 and 20’Q4 19’Q4 and 20’Q2
(it did not report
for 20’Q2) 40% 40%
40%

20% 20%
20%

0% 0%
0% Upper Base Lower Upper Base Lower Lower
Upper Base Lower case 1 case case 1 case 1 case case 1 case 2
case 1 case case 1
20 Q2 20 Q4 19 Q4 20 Q2 20 Q4
2020 Q4

French B2 MES weightings French B4 MES weightings Italian B1 MES weightings

80% 80% 80%

60% 60% 60%

40% 40% 40%

20% 20%
20%

0%
Upper Base Lower
0%
Upper Base Lower Lower
0%0%
case 1 case case 1 case 1 case case 1 case 2 Upper Base Lower
case 1 case case 1

19 Q4 20 Q2 20 Q4 19 Q4 20 Q2 20 Q4 20 Q2 20 Q4

24 |  EY insights on 2020 expected credit losses  June 2021

Insights on 2020 ECL disclosures
8.1 Overview
In 2020, banks provided a significant amount of
information on ECL and the level of detail surrounding
credit risk disclosures has generally improved. Although
this means that users were provided with a wealth of
information on credit risk and how it has evolved in the
financial year, consistency of the information disclosed
remains a challenge, especially considering the amount
of judgment involved in the application of IFRS 9. Despite
this challenge, IFRS 9 continues to be perceived by most
users as a definite improvement from IAS 39 and this was
especially tested under the exceptional circumstances of
the COVID-19 pandemic.
The key focus areas for the purpose of our analysis were
disclosures on:
• Multiple economic scenarios
• Post model adjustments and management overlays
• Sensitivity
• Vulnerable sectors
We have also looked at the granularity of information
provided on ECL, e.g., the breakdown of information
provided by portfolios (e.g., wholesale versus retail).
25 |  EY insights on 2020 expected credit losses  June 2021
8
The purpose of the analysis was to understand whether The length of the forecast period disclosed also varies,
the disclosures provided in those areas would allow a with some banks providing an outlook only for the next
good level of understanding of the ECL estimate and one or two years, whereas others projected further into
the underlying assumptions, as well as a comparison the future (up to five years).
between the banks.
The picture is quite mixed also on the number of the
key economic parameters disclosed (Figure 24). An
8.2 Multiple economic scenarios average of three to five parameters was disclosed,
One of the areas we have looked at is whether banks with some banks providing information only on one or
have provided the users of the financial statements with two parameters.
enough information to understand the key economic
Another challenge is with the presentation of those
assumptions behind the ECL estimates in the period.
parameters, with some banks providing yearly averages,
Overall, the level of detail is significant. Almost all banks others quarterly averages, others five-year averages
from the population that we analyzed disclosed key and others only the peak and trough values over the
information such as the number of economic scenarios forecast period.
used, the weightings assigned to these scenarios, the
key macro-economic parameters, as well as a narrative
to tell the story of how credit risk for the year has been
quantified. However, there was significant diversity in
practice on how the disclosures were provided.
In Figure 23, our analysis showed that the number of
economic scenarios disclosed differs across banks, with
the majority disclosing between three and four scenarios.
In general, UK banks disclosed more scenarios than other
European banks.
 Insights on 2020 ECL disclosures (continued) 8
Figure 23: Number of scenarios disclosed Figure 25: Probability weights assigned to the macroeconomic scenarios (for banks who disclosed it)

8 UK B1 15% 40% 40% 5%

7
UK B2 10% 30% 30% 30%
6
UK B4 10% 30% 40% 20%
No. of entities

4 UK B3 15.4% 15.5% 24.7% 24.2% 20.2%

3
ES B1 10% 25% 15% 45% 5%
2
ES B2 33% 34% 33%
1

0 IT B1 40% 55% 5%

1 2 3 4 5 6
NL B2 40% 50% 10%
No. of scenarios
NL B1 20% 60% 20%

FR B2 16% 50% 34%

FR B4 15% 10% 65% 10%

Figure 24: Number of parameters disclosed* CH B1 40% 60%

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%
5

4 Down-side 3 Down-side 2 Down-side 1 Base Upside 1 Upside 2

No. of entities

3
Finally, the weighting assigned to each of the economic Although all the above differences do not necessarily
2
scenarios varies significantly amongst banks in our mean that is never possible to compare banks against
1 population. For example, the weighting of the base each other, it appears evident that in most cases,
case varies from 24.7% to 65%, as shown in Figure 25 information has to be translated or adjusted before being
0
0 1 2 3 4 5 6 7 above. We have discussed in the previous sections (see compared and in some cases, this represents a challenge.
No. of parameters paragraph 7.3 on page 22) how this assumption can
significantly affect the overall ECL estimate.

* For the home country/key geographical area

26 |  EY insights on 2020 expected credit losses  June 2021

Insights on 2020 ECL disclosures (continued) 8
8.3 Post-model adjustments and Figure 26: Information disclosed on PMAs and MOs Figure 28: Is the impact of the base case disclosed in the
management overlays 8
sensitivity analysis?

Another area of focus of our benchmarking exercise 7

6

No. of entities
is the level of disclosures on post-model adjustments 5
(PMAs) and management overlays (MOs), i.e., any 4
No Yes
3
67% 33%
adjustment to the modeled output when estimating ECL. 2
As shown in Figure 26, the level of information provided 1
0
by banks varies, from providing no disclosures or only a PMAs or MOs Narrative Quantification Quantification
narrative disclosure, where the adjustment is disclosed not disclosed information of total impact of impact by
only portfolio
in nature but not quantified, to a more granular analysis,
with some banks even disclosing the impact of the
adjustment by portfolio and the effect on staging. Figure 27: Number of scenarios used in the sensitivity analysis 8.4 Sensitivity
It will be interesting to monitor the developments of 9 In 2020, banks continued to provide a significant level of
the judgments and estimates on this area in the next 8 multi-factor and single-factor sensitivity analyses
7
reporting periods and observe whether there is going to 6
(Figure 27) as well as narrative disclosures to further
No of entities

be a level of unwinding of the adjustments made to the 5 explain their approach and assumptions. UK banks
modeled loan loss allowance. It will depend on whether 4 generally provided sensitivities disclosures on more
3
the current economic uncertainty reduces and whether 2
economic scenarios than their European counterparts,
other events will occur in the future which may affect the 1 probably pushed by regulatory initiatives in this area,
models’ outputs. Disclosures in this area will therefore 0
1 2 3 4 5 such as DECL.1 The most common disclosure included
continue to be important. No. of scenarios three scenarios; however, certain inconsistencies were
found in the basis of preparation and presentation of the
relevant disclosures.
Often the sensitivity is done by weighting by 100%
some of the multiple economic scenarios, in which
case it is interesting to compare this to the effect of a
100% weighting of the base case, as it gives a measure
1 Disclosure of Expected Credit Losses, a Taskforce of preparers, investors
of the impact of multiple economic scenarios on the
and analysts jointly established and sponsored by the Financial Conduct booked ECL. As shown in Figure 28, we could find
Authority (FCA), the Financial Reporting Council (FRC) and the Prudential
Regulatory Authority (PRA).
this information for less than half of the banks in the
sensitivity disclosures.

27 |  EY insights on 2020 expected credit losses  June 2021

Insights on 2020 ECL disclosures (continued)
From our analysis, it was also observed that
approximately 40% of banks disclosed the effect of
sensitivities on staging (Figure 29), meanwhile, as shown
in Figure 30, less than one-third incorporated the effect
of PMAs and MOs. Fifteen per cent of banks disclosed
single-factor sensitivities (e.g., the impact of shifts in
the unemployment rate and house pricing indices on
retail portfolios, or moving all credit exposures to full
lifetime ECL).
The impact of the sensitivity analysis by portfolio (e.g.,
retail and wholesale) was disclosed in 56% of the cases,
with some banks providing more granularity than others
(the number of portfolios disclosed varies from two
to seven).
Figure 29: Is the effect of sensitivities on staging disclosed?
No Yes
61% 39%
28 |  EY insights on 2020 expected credit losses  June 2021
Figure 30: Is the impact of PMA/MOs considered in the
sensitivity analysis?
Yes
22%
Not
disclosed No
67% 11%
8.5 Vulnerable sectors
In 2020 we have seen banks providing more disclosures
of the industry sectors most significantly affected by the
COVID-19 pandemic, which are generally referred to as
“vulnerable sectors.” This is an area we understand is a
focus for many analysts in the current unprecedented
circumstances. The “top 5” sectors identified as
vulnerable, and for which banks in our sample have more
often identified a material level of exposures, are shown
in paragraph 6.2 on page 19.
As shown in Figure 31, almost all banks in the scope
of our benchmark provided a level of disclosure on
vulnerable sectors; however, as noted in section 6.2,
8
the information provided is not consistent across the
population. We have previously observed how differences
exist in the basis of presentation of the information.
Additionally, not all banks have disclosed their
off-balance sheet exposures to vulnerable sectors.
To provide more information on the quality of a bank’s
exposure to vulnerable sectors, it is generally considered
helpful to provide information about the staging of
these exposures. This disclosure was observed in only
27% of the cases, which suggests this could be an area
for improvement, although this is not a mandatory
disclosure under IFRS.
Figure 31: Information on vulnerable sectors disclosed
No
17%
Yes
83%
* Note that data is not fully comparable, as the basis of preparation differs
(e.g., gross vs. net book values or Exposures At Default) and, in some instances,
is not disclosed
Insights on 2020 ECL disclosures (continued)
Vulnerable sectors – Illustrative disclosure example
8
Figure 32: Vulnerable sectors — illustrative disclosure example

Staging allocation
Stage 1 Stage 2 Stage 3 Total
Values are for illustrative purposes only GCA ECL NCA GCA ECL NCA GCA ECL NCA GCA ECL NCA
Industry
Aviation 8 (2) 6 … … … … … … 15 (7) 8
Consider disclosing separately the
Leisure and entertainment 15 (2) 13 … … … … …
off-balance sheet exposures (e.g., … 43 (18) 25
Oil and gas 2 — 2 … … commitments)
loan … … … … 8 (3) 5
Retail 32 (2) 30 … … … … … … 78 (12) 66
Commercial real-estate 5 (1) 4 … … … … … … 16 (5) 11
Total 62 (7) 55 … … … … … … 160 (45) 115

Geographical distribution *
Values are for illustrative purposes only Region A Region B Region C Region D Region E NCA Total
Industry
Aviation 6 1 … … … 8
Leisure and entertainment 4 - … … … 25
Oil and gas 1 - … … … 5
Retail 18 12 … … … 66
Commercial real-estate 3 1 … … … 11
Total 32 14 … … … 115

44 IFRS 9 ECL, a benchmark for 2020 reporting disclosures and impacts

* Consider
NCA = Netdisclosing separately the GCA, ECL and NCA for the most significant geographical regions
Carrying Amount
GCA = Gross Carrying Amount
NCA = Net Carrying Amount

29 |  EY insights on 2020 expected credit losses  June 2021

Insights on 2020 ECL disclosures (continued)
Figure 33: Granularity of information
Retail vs. wholesale
No Yes
44% 56%
* For those who disclosed wholesale separately
8.6 Granularity of information
Overall, banks provided a significant level of information
on their credit risk exposures in their 2020 annual
financial statements, both in qualitative and quantitative
terms. In line with the general remark made in the
previous sections of the document, there still appears to
be room for an increased consistency of these disclosures
in order to allow for an easier comparison amongst
banks. For example, as shown in Figure 33 above, only
56% of the population in our sample provided the split
of their exposures between retail and wholesale for their
balances and movement tables. Of those who provided
such breakdown, although 80% disclosed mortgages
separately from their general retail exposures, only 30%
distinguished large corporates from SMEs.
30 |  EY insights on 2020 expected credit losses  June 2021
8
Stage 2 assets by risk grade, the granularity (i.e., number
of risk bands) is quite heterogeneous (Figure 35) and there
GCA and ECL breakdown is not always a clear and direct reconciliation between PDs
and internal or external ratings. This is, therefore, another
Large corporates vs. SMEs* Separate information on mortgages** area where further steps may be considered to enhance
the understanding by users of the information, although,
undoubtedly, there have been significant efforts made by
Yes No the banks in terms of transparency.
30% 20%
No Yes
70% 80% Figure 34: Split of COR by stage
, No
** For those who disclosed retail separately 28%
Yes
Almost all banks provided a disclosure of the overall cost 72%
of risk in the period by stage (i.e., the profit or loss ECL
charge, separately by Stage 1, 2 and 3), which is a metric
that many users of the financial statements, including
analysts, find particularly interesting (see Figure 34).
Another area of focus for many users of banks’ financial Figure 35: Breakdown of Stage 2 assets by risk grade
statements is the credit quality of the Stage 2 portfolio,
9 5 to 7
as it relates to exposures that can have significant 8
bands
differences in terms of probability of default and in 7
No of entities
overall quality. Providing further insights into the risks 6 8 to 10
bands
5
associated with those exposures is therefore considered 4
important. One piece of information that is generally 3 No
< 5 bands
disclosure
considered helpful in this area is the breakdown of the 2
Stage 2 population by risk grade (PD bands and/or 1
0
internal or external credit ratings). Although almost
all banks provide a disclosure of the breakdown of Format is heterogeneous (PDs vs ratings)
Mapping PDs/ratings often difficult
How EY teams can support you
Key considerations for IFRS 9 ECL
calculations in 2021 include assessing
model performance, incorporating 2020
data, planning for future unprecedented
economic events, remediating underlying
model improvements and governance over
releases of management overlays.
As we look ahead to IFRS 9 activities in 2021, the
current environment remains challenging for calculating
and explaining IFRS 9 expected credit loss, especially
as models and processes are still catching up with the
unprecedented environment and continued low default
levels. IFRS 9 also continues to be a focus area for
market analysts and banking regulators. The specific
areas of focus are:
• Updates to the macroeconomic outlook and scenario
weightings in light of frequent changes
• Determining the impact of support measures
and banks’ current visibility around the viability
of borrowers
• Improvements to data and monitoring processes to
assist in timely identification of troubled borrowers
• Understanding of model weaknesses to identify
in-model/post-model adjustments versus
model improvements
31 |  EY insights on 2020 expected credit losses  June 2021
• Balancing economic data and model results with
expert judgment of risk in portfolios
• Data and processes needed for transparency in
financial reporting
• Focus on the management of non-performing loans
• Control environment considerations considering
increased judgment and changes to processes
introduced, such as the introduction of government
support schemes
There are several distinct areas where we believe EY
can support your continued IFRS 9 improvement/
remediation efforts:
• Support in the remediation of underlying model
weaknesses that were exacerbated during
the pandemic
• Advise on the considerations of how to take into
account the 2020 economic data in modeling
frameworks. In 2020, the pandemic related shutdowns
and the unprecedented government stimulus
to consumers and businesses resulted in poor
performance of the loss forecasting models and banks
had to rely on multiple overlays and adjustments to
account for these effects. Banks will need to consider
impact of the data on model output, performance and
review and challenge
• Advise and support on how further unprecedented
economic shocks should be taken into account in
9
the IFRS 9 macro-economic models. The economic
environment over 2020 clearly demonstrated the
impact and significance of the key economic data
points have on the overall IFRS 9 ECL calculation
• Perform an assessment on the current in-model/post-
model adjustments processes, data and governance
and recommend improvements. Now that extensive
overlays have been used, banks need to consider
whether there is proper governance and sufficient
documentation and support for any releases of these
management overlays
Additionally, the approach to credit risk management
must be considered. The themes highlighted in this
publication demonstrate that the current assessment
process should be strengthened to incorporate the
financial ecosystem of counterparties. This will require
a tactical change in the immediate-term considering
methodologies, but the long-term approach can
leverage emerging technologies to ensure full alignment
to strategic objectives. This will strengthen the
resilience of banks and assist in identifying profits in a
restrictive economy.
At EY, we have developed several tools that can support
this analysis. The graphs produced in this publication
were created using “EY Spotlight.” This is a powerful
analytical tool that efficiently addresses the challenges
of IFRS 9 modeling in these unprecedented times by
providing deep insight about COVID-19 impacts on ECLs
at a glance.
EY contacts 10
EY teams have deep
Laure Guegan Rajeev Bansal Francesca Amatimaggio
experience of IFRS 9
implementation EY & Associés Ernst & Young LLP EY S.p.A.
+ 33 1 46 93 63 58 + 44 207 783 0898 + 39 027 221 22035
and understand the
laure.guegan@fr.ey.com rbansal@uk.ey.com francesca.amatimaggio@it.ey.com
complexities around
disclosures. We are
able to advise and Fabio Fabiani Sander de Ruiter Danny Buckley
support on augmenting Ernst & Young LLP EY Advisory Netherlands LLP Ernst & Young Chartered Accountants
your current process, + 44 20 7783 0816 + 31 6 290 83 992 + 353 1 479 2156
reflecting the guidance ffabiani@uk.ey.com sander.de.ruiter@nl.ey.com danny.buckley@ie.ey.com
of regulators and the
concerns of users. Tara Kengla Cassondra Polegri Paloma Muñoz
Your local EY contact Ernst & Young LLP Ernst & Young LLP Ernst & Young, S.L.
or the contacts + 44 20 7951 3054 + 44 20 7951 1873 + 34 60 626 6402
tkengla@uk.ey.com cpolegri@uk.ey.com paloma.munozgongora@es.ey.com
listed below will be
able to discuss your
requirements in Kevin Aycardo Michael Bosse Nikolas Stege
further detail. Ernst & Young LLP Ernst & Young GmbH Ernst & Young GmbH
+ 44 20 7951 5546 Wirtschaftsprüfungsgesellschaft Wirtschaftsprüfungsgesellschaft
kevin.aycardo@uk.ey.com + 49 511 8508 19642 + 49 511 8508 21509
michael.bosse@de.ey.com nikolas.stege@de.ey.com

Cédric Naud André Correia dos Santos

Ernst & Young Advisory Ernst & Young LLP
+ 33 1 46 93 67 40 + 44 20 7951 7064
cedric.naud@fr.ey.com asantos@uk.ey.com

32 |  EY insights on 2020 expected credit losses  June 2021

EY | Building a better working world
EY exists to build a better working world, helping create long-term value for
clients, people and society and build trust in the capital markets.
Enabled by data and technology, diverse EY teams in over 150 countries
provide trust through assurance and help clients grow, transform and
operate.
Working across assurance, consulting, law, strategy, tax and transactions,
EY teams ask better questions to find new answers for the complex issues
facing our world today.
EY refers to the global organization, and may refer to one or more, of the member firms of Ernst & Young
Global Limited, each of which is a separate legal entity. Ernst & Young Global Limited, a UK company limited
by guarantee, does not provide services to clients. Information about how EY collects and uses personal
data and a description of the rights individuals have under data protection legislation are available via
ey.com/privacy. EY member firms do not practice law where prohibited by local laws. For more information
about our organization, please visit ey.com.
© 2021 EYGM Limited
All Rights Reserved.
EYG no. 005319-21Gbl
EY-000133400-01.indd (UK) 06/21. Artwork by Creative Services Group London.
ED None
This material has been prepared for general informational purposes only and is not intended to be relied upon as accounting, tax, legal or
other professional advice. Please refer to your advisors for specific advice.
ey.com/uk