Expected cash flow estimation

advertisement
Riservato e confidenziale
Loan Portfolio Fair Value
Assessment
CDS’ experience from banking practice
Istanbul - May 2013
AGENDA
 Background
 Life time default estimation
-
Analytical framework
Main Results
 Expected cash flow estimation
-
Analytical framework
Discount rates
Early repayment rate
Portfolio pricing
 Benchmark analysis
 Conclusions
|2|
Strictly confidential 29/05/2013
WHY CONSIDER LOAN PORTFOLIO VALUATION?
Internal
securitization
Provision
estimation
Asset Disposal
Due diligence
Benchmarking

Banks are using these operations as a vehicle get funds from the ECB: the securitized
portfolio is eligible as collateral from ECB.

The complementary application of Life time cash flow estimation is the provision for
credit losses estimation. Provisions can be set on the basis of the current portfolio
status (default, not default) although many banks are starting to determine provisions
also on the basis of estimation.

This is a classical method that Banks and Financial houses use to get funds. The
crisis has created a very interesting disposal market fed by companies specialized in
debt collection.

Expected cash flows can be used to review the credit portfolio governance and
management, and get preliminary estimates of pricing and profitability of the
portfolio, in accordance with Group portfolio management strategy (i.e. run-off)

Banks, especially at the Headquarter level, need to benchmark a loan portfolio’s
lifetime estimation versus the system, aiming to verify the robustness of the
estimation, or to highlight particular characteristics of the managed portfolio.
|3|
3
Strictly confidential 29/05/2013
CASE STUDY: BUSINESS NEEDS

Internal
securitization
Provision
estimation
Asset Disposal
Due diligence
Benchmarking
The Bank requested to CRIF to support them in a value assessment of their
consumer loan portfolio, prior to an internal securitization operation.

The complementary application of Life time cash flow estimation is the provision for
credit losses estimation. Provisions can be set on the basis of the current portfolio
status (default, not default) although many banks are starting to determine provisions
also on the basis of estimation.

This is a classical method that Banks and Financial houses use to get funds. The
crisis has created a very interesting disposal market fed by companies specialized in
debt collection.

Expected cash flows can be used to review the credit portfolio governance and
management, and get preliminary estimates of pricing and profitability of the
portfolio, in accordance with Group portfolio management strategy (i.e. run-off)

Banks, especially at the Headquarter level, need to benchmark a loan portfolio’s
lifetime estimation versus the system, aiming to verify the robustness of the
estimation, or to highlight particular characteristics of the managed portfolio.
|4|
4
Strictly confidential 29/05/2013
PROJECT APPROACH
Project objectives
Project activities
1
Life time default
(LTD) estimation
Customer: Italian subsidiary of leading
international banking group

In order to estimate the life time risk on
the securitized portfolio, CRIF defined a
framework that, through the Markov chain
approach, makes possible the LTD estimation
on a time window of 10 years.

In order to estimate the securitized portfolio
value we estimated the future cash flows
splitting the portfolio by product (personal
loans, consumer loans and auto loans) and
level of risk (tranching).

The specific LTD, LGD and early repayment
rate estimations together with the discount
rates have contributed to determine the
future cash flow estimation
Main goal:
 Calculating portfolio’s expected
value & pricing before undertaking
the decision if running or not the
securitization (by the HQ)
2
Expected cash
flow estimation
 Assessing the impact of the cost of
funding of the bank into the
expected value of the portfolio
 Identifying high/low performing
tranches of the portfolio, leveraging
even on external data (CRIF Credit
Bureau)
3
|5|
Benchmarking
parameter
estimation
5

To build similar portfolios based on generic
samples retrieved from the Italian Credit
Bureau. On the basis of these samples we
computed the benchmarking LTD curves.
Strictly confidential 29/05/2013
AGENDA
 Background
 Life time default estimation
-
Analytical framework
Main Results
 Expected cash flow estimation
-
Analytical framework
Discount rates
Early repayment rate
Portfolio pricing
 Benchmark analysis
 Conclusions
|6|
Strictly confidential 29/05/2013
LIFE TIME DEFAULT ESTIMATION: ANALYTICAL FRAMEWORK

The Life Time Default (LTD) estimation is part of the multiperiod estimation problems, in
which the evolution from an initial state to a final state passes through "n" intermediate
observable states, precisely, over several periods. The technique used by CRIF is based on
"Markov chains".

Credit scoring uses snapshots of historical information, observations and outcomes, to develop a
risk-ranking tool. For behavioral-risk scoring, most models will use a one-year outcome. Then to
determine the probability of an account going bad, or defaulting, over any given period, historical
information is again used to work out the rates. But what if percentages are needed within, or
beyond, the one-year period, or whatever period which was used for the scorecard?

Markov chain allows the business to predict the future distribution (default/non default)
using only the current distribution (without default) and a transition matrix indicating the
expected movements between states.
Risk
Multi period frame
|7|
7
Strictly confidential 29/05/2013
LIFE TIME DEFAULT ESTIMATION-ANALITICAL FRAMEWORK
1
Objective
2
Methodological
foundations

Starting from the reference date of December 2012 the objective was of estimating LTD
on the outstanding loans present at that time in the “customer” portfolio.

The base of Markov chain estimation consists of building transition matrices that, on the
basis of the past experience, can be used in order to extrapolate the future risk

The life time risk extrapolation is based on two simple assumptions:
 When a contract is classified as defaulted do not change the status any more.
 The probability to change status along the life time is constant and the transition
represents these probabilities.
3
Data sources
matrix

Therefore the simple iteration of the transition matrix per the number of periods that
cover the whole residual life of the securitized loan portfolio is able to estimate the LTD.

For the “customer” were developed two transition processes able to estimate the marginal
default probabilities over a time frame of 10 years (120 months) both for the personal
loans and the auto loan portfolios.

The analysis was realized on the loan portfolios provided by the “customer” , who
provided the payment history of each loan useful to set the transition matrices on the
basis of an internal default definition.

CRIF has integrated the “customer” historical loan portfolios with Credit Bureau Score in
order to define a robust risk tranching and coherent with the benchmarking activity.

The definition of default has been established on the basis of the following two variables:
 Risk Class (number of insolvency in each period)
 Contract Status (issued, closed, early pay off, write off, into areas, DBT)
|8|
Strictly confidential 29/05/2013
LIFE TIME DEFAULT ESTIMATION-ANALYSIS AND DELIVERABLES
Performed Analysis
Deliverables
Default rate
 Initially, it is analyzed the distribution of Perform
Not defaulted personal loan portfolio at 06/2012 with performance at 12/2012
with respect the default rate defined as the
target variable
Bonis
Crif Bureau
Score
 The default status and the Credit Bureau Score
tranches represented the possible moving
statutes of each contract
Trantion matrix
%
#
1,077
5.93%
372
745
4.11%
104
initial peaks more pronounced for riskier
classes.
|9|
1,449
7.60%
25.67%
66.58
11.34%
849
4.45%
12.25%
7.36
0.65
C
(D,E)
1,982
10.92%
126
13.74%
2,108
11.06%
5.98%
D
(F,G)
749
4.13%
33
3.60%
782
4.10%
4.22%
0.07
E
(H,I)
4,640
25.57%
108
11.78%
4,748
24.91%
2.27%
10.69
F
(L,M,N,O)
7,361
40.56%
149
16.25%
7,510
39.39%
1.98%
22.25
G
(P)
1,593
8.78%
25
2.73%
1,618
8.49%
1.55%
7.07
18,147
100%
917
100%
19,064
100%
4.81%
114.67
Transition June 2012 – December 2012
Default
A
B
C
D
E
F
G
Totale
A
B
C
D
E
F
G
25.7%
12.2%
6.0%
4.2%
2.3%
2.0%
1.5%
55.9%
23.1%
9.3%
2.8%
2.4%
1.0%
0.3%
7.5%
33.5%
11.3%
3.7%
2.3%
1.0%
0.7%
9.9%
25.4%
52.1%
8.6%
4.0%
1.6%
0.5%
0.2%
1.8%
7.2%
29.7%
4.2%
0.8%
0.5%
0.3%
2.4%
5.4%
37.2%
57.6%
11.7%
2.2%
0.4%
1.4%
7.6%
11.5%
26.8%
72.5%
29.2%
0.1%
0.2%
1.0%
2.3%
0.4%
9.4%
65.1%
100.0%
100.0%
100.0%
100.0%
100.0%
100.0%
100.0%
1,449
849
2,108
782
4,748
7,510
1,618
 The
 The curve shows a decreasing trend with an
40.57%
(A,B)
(C)
PP
a tranche “A” contract in June 2012 has a 25.7%
probability of going after six months in default, has a
55.9% probability of staying in tranche “A”
additional loans at that time reaches the default
status.
IV
%
Personal loan portfolio transition matrix (06/2012 – 12/2012)
loans portfolio is reported in the table. The
matrix represents the transition probabilities of
a contract on score tranches or the default
status between two periods:
 Each point represents the percentage of
Default rate
#
B
 The transition matrix of the performing personal
two graphs show the evolution of marginal
probabilities of default for each risk class,
Total
%
A
Total
Rsik measurement
Default
#
LTD curves
Personal Loans
30%
20%
10%
10%
5%
0%
giu-12dec12 jun13 dec13 jun14 dec14 jun15 dec15 jun16 dec16 jun17 dec17 jun18 dec18 jun19 dec19
A
B
C
D
E
F
AutoLoans
15%
G
0%
giu-12dec12 jun13 dec13 jun14 dec14 jun15 dec15 jun16 dec16 jun17 dec17 jun18 dec18 jun19 dec19
A
B
C
D
E
F
Strictly confidential 29/05/2013
AGENDA
 Background
 Life time default estimation
-
Analytical framework
Main Results
 Expected cash flow estimation
-
Analytical framework
Discount rates
Early repayment rate
Portfolio pricing
 Benchmark analysis
 Conclusions
| 10 |
Strictly confidential 29/05/2013
EXPECTED CASH FLOWS ESTIMATION-ANALITICAL FRAMEWORK

To estimate the expected cash flows is necessary to build the amortization schedule of the loans from the
evaluation date until the date of expiration, based on the amount of the installment, the residual debt and the
annual rate (APR) provided by the bank on individual basis.

Once it were obtained the values of the residual capital and of the interest at any time t, it has been
developed a binomial tree where at each subsequent time t (t = 1,2... until expiration), the probability of
being current or default status is derived from the Markov process.

In case of default at the time t, the cash flows are represented by the installment collected up to t-1, with
the addition of the recovered residual debt plus the installment gained in the period, all multiplied by the
recovery rate:
Revenuest =
Installments
+
the amount of the installments
already paid from 0 to s =t-1

(1-LGD) * (Installmentt +Residual Debtt )
current installment+the
recovered residual debt
at the default event.
In case of not default at the time t, revenues are represented by the sum of installments paid up to s = t-1,
with in addition the installment of the period t:
Revenuest = Installments + Installmentt
Expected cash flows is determined by the sum of the discounted expected
revenues from the evaluation date until the date of expiration
| 11 |
Strictly confidential 29/05/2013
DISCOUNT RATES AND EARLY REPAYMENT RATE
Early repayment rates
Discounts rates

Revenues resulting from the application of the
algorithm are then discounted at each time t, In
order to give an expected value of the portfolio that
takes into account the cost of funding of the bank.

For the evaluation of the expected cash flows, it was
necessary to define the early repayment rate of a
credit line,.

The repayment rate was measured, on annual basis,
in the year 2012. The ratio is measured as the
number of contracts that expire in advance during
2012, with respect to the number of contracts alive
at the beginning of that year.

The early repayment rates are introduced into the
cash flows analysis: grater is the repayment rate
lower will be the expected cash flows.

Here we can see the early repayment rates chose for
the Personal Loans portfolio.
Liquidity curve = risk free rate + liquidity spread
Greater is the cost of funding lower will
be the expected value of the portfolio
Upper Bound
Lower Bound
| 12 |
Personal Loans Early repayment rate grouped by
Perform tranches:
Rate f or PP tranche A
0.62%
Rate f or PP tranche B
1.65%
Rate f or PP tranche C
4.42%
Rate f or PP tranche D
10.66%
Rate f or PP tranche E
14.49%
Rate f or PP tranche F
11.27%
Rate f or PP tranche G
9.97%
Strictly confidential 29/05/2013
PORTFOLIO PRICING
 Portfolio pricing is determined by the ratio of the expected cash flows to the total balance.
 As for bonds, portfolio priced over 100 are more profitable meanwhile portfolio priced under
100 are less profitable for the originators.
 The
tranching of the portfolios is defined on the basis of the CRIF Credit Bureau Score.
The price is mainly determined by the risk profile and the structure of the amortization plan
Pricing = Expected C.F./Total Balance
>100
OK
=100
<100
| 13 |
KO
Perform
Pricing
A
98,2%
B
103,1%
C
105,8%
D
108,5%
E
108,9%
F
109,6%
G
106,3%
Total
108,0%
Strictly confidential 29/05/2013
AGENDA
 Background
 Life time default estimation
-
Analytical framework
Main Results
 Expected cash flow estimation
-
Analytical framework
Discount rates
Early repayment rate
Portfolio pricing
 Benchmark analysis
 Conclusions
| 14 |
Strictly confidential 29/05/2013
BENCHMARKING ANALYSIS
Comparison of Default Rate
 To compare and validate the results obtained on specific “customer” portfolios, CDS
has developed a benchmarking analysis extracting two random samples, from
Italian Credit Bureau, that has similar features to those of “customer” .
 The main purpose of the benchmarking analysis was to compare the risk detected
in the Credit Bureau samples, to verify if the behavior and the structure of the
“customer” portfolios are aligned or not with what is observed in the total market.
 As can be seen from the graphs, the risk curves by Credit Bureau Score bands that
are obtained from the samples of the system show comparable trends to those
defined on the portfolios of the customer
| 15 |
Strictly confidential 29/05/2013
BENCHMARKING ANALYSIS
Customer and Credit Bureaus LTD curves comparison
Credit Bureau sample
Customer sample
Personal loan
30%
20%
10%
0%
giu- dec12 jun13 dec13 jun14 dec14 jun15 dec15 jun16 dec16 jun17 dec17 jun18 dec18 jun19 dec19
12
Auto & consumer
loans
A
B
C
D
E
F
G
20%
10%
0%
giu-12 dec12 jun13 dec13 jun14 dec14 jun15 dec15 jun16 dec16 jun17 dec17 jun18 dec18 jun19 dec19
A
| 16 |
B
C
D
E
F
Strictly confidential 29/05/2013
AGENDA
 Background
 Life time default estimation
-
Analytical framework
Main Results
 Expected cash flow estimation
-
Analytical framework
Discount rates
Early repayment rate
Portfolio pricing
 Benchmark analysis
 Conclusions
| 17 |
Strictly confidential 29/05/2013
CONCLUSIONS
Internal
securitization

CRIF can support banks in a preliminary assessment phase in order to identify the best
strategy for securitization.
Provision
estimation

CRIF can help banks define the amount of provisions based on the expected
lifetime risk or limited to fixed periods (e.g. 36 months), as required by the new IAS
Regulation.

CRIF can support banks in an evaluation phase of the portfolio or specific segments to
handle the negotiations with potential buyers.

CRIF, on the basis of the expected cash flows estimation, can support Banks in
reviewing the credit portfolio governance and management, and get preliminary
estimates of pricing and profitability of the portfolio.

CRIF, where the Credit Bureau information is available, is able to compare the expected
risk (life time or multi-period) of the Bank with those of the system, providing evidence
of benchmarking useful in all areas listed above but, more generally, to assess
whether the value of its portfolio is determined by a systemic context or from a
specific credit management approach.
Asset Disposal
Due diligence
Benchmarking
| 18 |
18
Strictly confidential 29/05/2013
Crif Decision Solutions
Via M. Fantin 1-3
40131 Bologna
Tel.: + 39 051 4176111
Fax.: + 39 051 4176010
www.crif.com
Riservato e confidenziale
Business Consulting – R&I Innovation
Download