6

Question 1

Question 1
When modeling LGD, we focus on
a) operational defaults.
b) technical defaults.
c) real defaults.
d) all of the above.

Answer 1

c) real defaults.

Question 2

Question 2
When modeling LGD, it is of key importance that the default definition used is
a) the same as for PD.
b) different from the PD default definition.

Answer 2

a) the same as for PD.

Question 3

Question 3
Changing the default definition impacts
a) only the PD.
b) only the LGD.
c) both the PD and LGD.

Answer 3

c) both the PD and LGD.

Question 4

Question 4
LGD focusses on
a) economic loss.
b) accounting loss.

Answer 4

a) economic loss.

Question 5

Question 5
The most popular way to measure LGD is the:
a) Workout approach.
b) Market approach.
c) Implied historical approach.
d) Implied market approach.

Answer 5

a) Workout approach.

Question 6

Question 6
When modeling LGD, we need
a) 5 years of data for retail exposures and 5 years of data for wholesale exposures.
b) 7 years of data for retail exposures and 5 years of data for wholesale exposures.
c) 5 years of data for retail exposures and 7 years of data for wholesale exposures.
d) 7 years of data for retail exposures and 7 years of data for wholesale exposures.

Answer 6

c) 5 years of data for retail exposures and 7 years of data for wholesale exposures.

Question 7

Question 7
When measuring LGD, the vintage approach looks at
a) default data instead of loan origination date.
b) loan origination date instead of default date.

Answer 7

b) loan origination date instead of default date.

(2) vintage approach  loan origination date
(1) Static pool/ cohort approach  time of default

Question 8

Question 8
When modeling LGD, incomplete workouts
a) should be taken into account.
b) can be ignored.

Answer 8

a) should be taken into account.

Question 9

Question 9
Negative LGDs should be
a) left as is.
b) capped at zero.
c) set to 1.
d) ignored.

Answer 9

b) capped at zero.

Question 10

Question 10
When modeling LGD, indirect costs should be
a) considered for IFRS 9 but not for Basel.
b) considered for both Basel and IFRS 9.
c) considered for Basel but not for IFRS 9.
d) unconsidered for both Basel and IFRS 9.

Answer 10

c) considered for Basel but not for IFRS 9.

Question 11

Question 11
Which of the following should not be included in an LGD data set:
a) Loan characteristics.
b) Borrower characteristics.
c) Data about the non-defaulters.
d) Macro-economic factors.

Answer 11

c) Data about the non-defaulters.

Borrower characteristics
Macro-economic factors
Loan characteristics
* Country-related features

Question 12

Question 12
A higher Loan-To-Value (LTV) ratio typically gives a
a) Higher LGD.
b) Lower LGD.

Answer 12

a) Higher LGD.

loan to value (LTV) = ratio of the value of loan (= exposure) to value of underlying asset

Question 13

Question 13
Which of the following measures cannot be used to make the splitting decision in a regression tree?
a) Mean Squared Error (MSE).
b) ANOVA/F-test.
c) Entropy.

Answer 13

c) Entropy.

Question 14

Question 14
The LGD target variable can typically be approximated by a
a) Normal distribution.
b) Beta distribution.
c) Binomial Distribution.
d) Poisson distribution.

Answer 14

b) Beta distribution.

Question 15

Question 15
Zero-one-inflated Beta regression can be used
a) in case the LGD can equal 0 or 1.
b) only when the LGD is always strictly bigger than 0 and strictly less than 1.

Answer 15

a) in case the LGD can equal 0 or 1.

Question 16

Question 16
Logistic regression
a) can never be used for LGD modeling.
b) can be used for LGD modeling.

Answer 16

b) can be used for LGD modeling.

Question 17

Question 17
Two stage models for LGD decompose the problem into
a) a regression problem to model cures and a classification problem to model losses.
b) a classification problem to model cures and a classification problem to model losses.
c) a classification problem to model cures and a regression problem to model losses.
d) a regression problem to model cures and a regression problem to model losses.

Answer 17

c) a classification problem to model cures and a regression problem to model losses.

Question 18

Question 18
When modeling LGD, the R-squared value
a) is typically rather high since LGD is quite easy to predict.
b) is typically rather low since LGD is quite difficult to predict.

Answer 18

b) is typically rather low since LGD is quite difficult to predict.

Question 19

Question 19
According to the Basel Accord, the LGD cannot be less than
a) the long-run time-weighted average loss rate given default.
b) the long-run default-weighted average loss rate given default.
c) the long-run exposure-weighted average loss rate given default.

Answer 19

b) the long-run default-weighted average loss rate given default.

Question 20

Question 20
In case there is no PD-LGD correlation, we can use the
a) average LGD.
b) downturn LGD.
c) upturn LGD.
d) all of the above.

Answer 20

a) average LGD.

Question 21

Question 21
According to the Basel Accord, a bank must estimate an LGD for each facility that aims to reflect
a) economic upturn conditions.
b) average economic conditions.
c) economic downturn conditions.

Answer 21

c) economic downturn conditions.

Question 22

Question 22
Which statement is NOT CORRECT?
a) For on-balance sheet exposures, such as term loans, installment loans and mortgages, the EAD is defined as the nominal, outstanding balance, net of specific provisions.
b) The EAD decreases as the loan matures and it needs to re-considered at the time of capital calculation.
c) When defining EAD, there can be downward adjustments applied for amortization or expected prepayments.
d) When defining EAD for off balance sheet exposures, such as credit cards and credit lines, you need to take into account what portion of the undrawn amount is likely to be converted into credit upon default.

Answer 22

c) When defining EAD, there can be downward adjustments applied for amortization or expected prepayments.

Question 23

Question 23
When modeling EAD, Basel uses the
a) Credit Conversion Factor (CCF) approach.
b) Credit Equivalent (CEQ) approach.
c) Limit Conversion Factor (LCF) / Loan Equivalent (LEQ) approach.
d) Used Amount Conversion Factor (UACF) approach.

Answer 23

a)ccf??

Question 24

Question 24
Assume a credit limit of $2500, an EAD of $1000 and a drawn amount of $500
The CCF then equals
a) 0%
b) 25%
c) 50%
d) 100%

Answer 24

b)25%

CCF=(EAD-Drawn)/(Limit-Drawn)
(1000-500)/(2500-500)

Question 25

Question 25 Conservative estimates of EAD set the CCF to a) 0 b) 100%

Answer 25

b) 100% ## Footnote The higher the CCF, the higher the expected exposure in the event of default

Question 26

Question 26 Which statement is NOT CORRECT? a) The EU and US regulation also introduce the idea of a margin of conservatism and economic downturn EAD if it turns out that the EAD is volatile over the economic cycle. b) When modeling EAD, also the data observation period is identical to LGD with five years for retail exposures and seven years for corporate exposures. c) Additional drawings prior to default can either be included in LGD or EAD d) A downturn EAD is needed for IFRS 9.

Answer 26

d) A downturn EAD is needed for IFRS 9. ## Footnote In IFRS 9: no downturn EAD needed

Question 27

Question 27 Drawings post default can be included a) only in LGD. b) only in EAD. c) in LGD or EAD.

Answer 27

*c) both * ## Footnote – both EAD and LGD dependent on length of recovery process – EAD fixed at time of default, LGD dependent on length of recovery process

Question 28

Question 28 Which statement about EAD modeling is NOT CORRECT? a) The cohort method groups defaulted facilities into discrete calendar periods (for example, 12 months unless another time period is more conservative and appropriate) according to the date of default. b) The fixed horizon method collects information about risk factors and drawn/undrawn amount for a fixed time interval prior to the date of default (at least 12 months unless another time period is more conservative and appropriate) and the drawn amount on the date of default, regardless of the actual calendar date on which the default occurred. c) The variable time horizon approach is a variant of the fixed time horizon approach using several reference times within the chosen time horizon. d) The momentum method defines EAD as EAD=Drawn+CCF.(Limit-Drawn).

Answer 28

d) The momentum method defines EAD as EAD=Drawn+CCF.(Limit-Drawn). ## Footnote – work with EAD = LCF.Limit (LCF = Limit Conversion Factor)

Question 29

Question 29 For prediction purposes, it is recommended to a) cap a negative CCF to 0. b) leave a negative CCF untreated.

Answer 29

a) ## Footnote – CCF can be negative when borrower has paid back portion of debt prior to default – cap to 0 in data for estimation – predicted CCFs cannot be negative

Question 30

Question 30 CCF modeling is very similar to: a) PD modeling b) LGD modeling

Answer 30

b) LGD modeling ## Footnote In summary, LGD focuses on the potential loss, PD on the likelihood of default, and CCF on the proportion of exposure drawn down in the event of default. These parameters are essential for assessing and managing credit risk in financial institutions. CCF is concerned with estimating the proportion of an exposure that is expected to be drawn down in the event of default. LGD, on the other hand, focuses on the proportion of the exposure that is not recovered in the event of default.

Question 31

Question 31 Which of the following variables can be used to predict CCF? a) Type of obligor b) Obligor’s access to other sources of funding c) Factors affecting borrower’s demand for funding/credit facilities d) Expected growth in obligor’s business e) Utilization (drawn/limit) f) Characteristics of facility g) Credit risk h) all of the above.

Answer 31

h) all of the above.

Question 32

Question 32 The predictive performance (as, e.g., measured by the R-squared, MSE, MAD) for predicting CCF is usually a) rather high. b) rather low.

Answer 32

b) rather low

Question 33

Question 33 Which statement is NOT CORRECT? a) The correlation of PDs across obligors is captured by the asset correlation factor in the Basel Accord. b) PD and LGD are treated as dependent in the Basel Accords. c) A positive correlation between PD and LGD is often found. d) PD, LGD and EAD correlations are important to be aware of during both modeling, validation and stress testing.

Answer 33

b) PD and LGD are treated as dependent in the Basel Accords.