TopQuants: Modelling Liquidity Risk
Mortgage example
Amsterdam, 12 November 2014
Outline of this session
Introduction to liquidity risk
Behavioural modelling
Stress testing
Mortgages
1 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Introduction to liquidity risk
What is liquidity risk?
Different definitions of liquidity risk:
• The risk that a given security or asset cannot be traded quickly enough in the market to prevent a loss (limited
scope);
• The risk that a financial firm is not able to meet its debt obligations without incurring unacceptably large losses
(limited scope);
• The risk that an organisation is not able to meet its cumulative net cash outflow over a specific period of time.
Important aspects of liquidity:
• Cumulative net cash outflow, so both cash in and outflow;
• Specific period;
• Ability to convert assets into cash.
Possible causes of liquidity risk:
• Bank run;
• Prepayment of mortgage;
• Margin calls.
2 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Introduction to liquidity risk
What is the difference between solvency and liquidity?
There is a distinct difference between solvency and liquidity.
Solvency Liquidity
The degree to which the current assets of an The degree to which an organisation is able to meet
individual or entity exceed the current liabilities of that its cumulative net cash outflow over a specific period
individual or entity. of time.
Examples of mutual influence
• A low level of solvency makes it difficult to attract credit facilities;
• A low level of solvency increases the chances of a bank run;
• Fire sale of assets to meet liquidity need leads to a reduction of
asset prices.
3 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Introduction to liquidity risk
Why do we need to focus on this?
Unexpected and material liquidity outflows both under normal and stressed market conditions are serious threats for
banks.
In recent years we observed three major trends that effects how liquidity risk is perceived by banks:
• Rising funding costs: The credit/liquidity crunch forced banks to reassess their transfer pricing policies in light
of increased costs of funding and occasional shortage of liquidity in the market;
• More stringent regulatory requirements: Increased and more stringent regulation (e.g. in UK and The
Netherlands) forces banks to increase their focus on liquidity risk;
• Need for profitability improvement: As costs are increasing there is a need to allocate liquidity costs more
efficient.
Until now focus within liquidity modelling was mainly on contractual maturities. Caused by these trends the
behavioural maturity becomes more important.
4 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Behavioural Modelling
Objective and purpose
The objective is to develop models that are fit for purpose:
• Funding: input for funding plan;
• Liquidity stress testing: input for liquidity stress testing framework (survival period);
• Pricing: input for direct and/or indirect liquidity costs.
The final outcome of a behavioural model (example for savings):
Difference contractual and behavioural maturity
12000
10000
Amount in EUR
8000
6000
4000
2000
0
now < 1 week < 1 month< 3 month< 6 month< 9 month < 1 year < 2 year < 3 year < 4 year < 5 year < 10 year
Maturity buckets
Contractual maturity Behavioural maturity
5 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Behavioural Modelling
Scope
For a full overview of liquidity risk, all balance sheet items need to be taken into account.
• For liabilities the main risk is early withdrawal;
• For assets the main risk is that a loan is extended (rolled over) after the maturity date or prepaid before the
maturity date.
In general two types of models can be used:
• For maturing products: (multinomial) logistic regression;
• For non-maturing products: combination of logistic regression (survival) and an ARDL model (avg. balance).
Assets Liabilities
Mortgages Term Deposits
Maturing products
Term Loans
Non-Maturing
Current accounts Savings
products
6 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Liquidity stress testing
7 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Why liquidity stress testing?
Regulatory perspective
• Identify the risk that a bank is not able to meet it’s net cash flow output;
• Liquidity buffer is used to dampen the effect of unexpected cash outflows;
• Stress testing (ILAAP) should take into account stress scenario’s that:
• Last at least 3 months;
• Unsecured funding is not rolled over during 3 months;
• Retail outflow doubles to at least 20% during the first month;
• Facilities are drawn for 50%;
• Haircuts rise;
• Market value of liquid assets goes down;
• FX markets close for 2 weeks;
• 2 notch downgrade within 3 months.
8 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Maturity calendar – net cash-flows
• Behavioural cash flows are based on contractual cash flows and modified for behavioural aspects;
• Based on estimated CPR’s and similar metrics, a cash flow calendar can be constructed for a liquidity model;
• Typically separate models are developed for each balance sheet item and interactions are not taken into
account.
Stressed Survival period Survival period
100%
80%
60%
Liquidity buffer 40%
20%
0%
-20%
-40%
-60%
-80%
-100%
Base scenario Stress
9 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Aggregating different balance sheet models
• Typically each balance sheet segment has it’s own characteristics and model;
• Challenge is how to correctly take into account diversification effects.
Interest rates down scenario
100%
80%
60%
40%
20%
0%
-20% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
-40%
-60%
-80%
-100%
Assets - BaU Assets - Stressed Liabilities - BaU Liabilities - Stressed
10 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
VAR approach for residual aggregation
To aggregate liquidity models, a VAR approach can be used
• Interdependence of these residual series can be estimated using a vector autoregressive model or VAR(p)
model:
𝑌𝑡 = 𝛼 + 𝑌𝑡−1 Φ𝑗 + 𝑈𝑡 , 𝑈𝑡 ~𝐼𝐼𝐷 0, Σ or
𝑗=1
𝑝 𝑚
𝑦𝑡𝑖 = 𝛼𝑖 + 𝑦𝑡−𝑗,𝑘 𝜙𝑗,𝑘𝑖 + 𝑢𝑡𝑖
𝑗=1 𝑘=1
• Which can be estimated by OLS given normality of the residuals, where Σ can be estimated as follows:
𝑛
1
Σ= 𝑈𝑡′ 𝑈𝑡
𝑛
𝑡=1
• The VAR estimates can be used to perform a Monte Carlo simulation that can provide quantile estimates of the
maturity calendar.
11 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Mortgages
Example
12 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Three options define the behavioral cash flows of a Mortgage
Definition liquidity risk: Deviation from contractual cash flows.
Every period we modelled three possibilities that could influence the behavioural cash flows:
• Curtailment (partial prepayment), part of the outstanding is repaid;
• Conversion, the loan(part) is closed and replaced by a new loan(part);
• Full Prepayment, the loan is fully repaid.
Conversion
100% 100%
Prepayment
Curtailment
0% 0%
2014 2019 2024 2029 2034 2014 2019 2024 2029 2034
Contractual Behavioural Contractual Behavioural
13
To model the three different options a multinomial logit model is used
The multinomial logit model uses multiple logistic regressions to estimate the parameters.
The response variable in our case is:
1: full prepayment
2: curtailment
𝑦= 3: conversion
4: no changes
Based on this model the probabilities are defines as follows (j=1..3):
𝑒 𝛼𝑗 +𝛽𝑗 𝑋
𝜋𝑗 = 𝑃 𝑦 = 𝑗 =
1 + 3𝑗=1 𝑒 𝛼𝑗 +𝛽𝑗 𝑋
And for the last category:
1
𝜋𝑘 = 𝑃 𝑦 = 4 = 3 𝛼𝑗 +𝛽𝑗 𝑋
1+ 𝑗=1 𝑒
This produces a set of probabilities that sums to one:
4
𝜋𝑖 = 1
𝑖=1
14
Possible risk drivers for Mortgages Loans
Unemployment Loan Age
Macro-economic risk drivers rate
House prices Amortization
To model liquidity type
risk for mortgages
two types of risk
drivers are used
Mortgage Seasonality Contract specific risk drivers
rate
Interest rate Type of
term collateral
15 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
Model fit - Testing
Residual testing
Residual testing is performed on aggregated level:
𝜋𝑖,𝑡
𝑙𝑛 = 𝑋𝑖,𝑡 𝛽𝑖,𝑡 + 𝜖𝑖,𝑡
𝜋𝑘,𝑡
• Assuming the last response ‘no changes’ (k) has a probability close to one, for 𝑖 ≠ 𝑘:
𝜋𝑘 ≈ 1 − 𝜋𝑖
• denoting 𝜋𝑖 as the realized fraction of an event, it follows that
𝜋𝑖,𝑡 𝜋𝑖,𝑡
𝜖𝑖∗ = 𝑙𝑛 − 𝑙𝑛 ~𝑁𝐼𝐷(0, 𝜎𝑖 )
1 − 𝜋𝑖,𝑡 1 − 𝜋𝑖,𝑡
4% 1.0
Prepayment rate
3% 0.5
2% 0.0
2011 2012 2013 2014
1% -0.5
0% -1.0
2011 2012 2013 2014
Predicted Realized Residual
16 TopQuants - Liquidity Risk © 2014 Deloitte The Netherlands
From model output to behavioural calendars
The four probabilities are used to perform a behavioural cash flow projection
• Use a Monte Carlo Simulation to simulate paths based on the output probabilities:
• Use the probabilities to define a Single Monthly Mortality rates to project the cash flows.
𝑘
𝑆𝑀𝑀 = 𝜋𝑖 ⋅ 𝑣𝑖
𝑖=1
πi is the probability that event j will occur and vi contains the mortality rate per event.
Contractual cash flows have to be subtracted.
SMM
Outstanding
SMM
Time
17
Results
Outstanding over time Distribution Weighted Average Life Cash flow calendar
Behavioural and contractual Weighted average life: Cash flows contractual and
outstanding. behavioural
𝑁
𝑡=1 𝐶𝐹𝑡 ∗ 𝑡
𝑊𝐴𝐿 = 𝑁
𝑡=1 𝐶𝐹𝑡
6000 700
1,000,000,000
900,000,000 5000 600
800,000,000
Cash flow (mio)
500
Observations
700,000,000 4000
600,000,000 400
500,000,000 3000
400,000,000 300
300,000,000 2000
200
200,000,000
1000
100,000,000 100
0 0 0
2035
2014
2017
2020
2023
2026
2029
2032
2038
2041
2044
1 5 9 13 17 21 25 29
Behavioural
N
Contractual Behavioural Contractual
18
Modelling challenges
Challenges in model estimation:
• Limited data availability: Liquidity models are first generation models - not all desired data available;
• Regulatory changes influence client behaviour;
• Historical changes:
− Mortgage types;
− Portfolio changes.
Challenges in cash flow projection:
• Limited historical data with stress;
• Limited availability of benchmarks.
19
Contact details
Bauke Maarse Maurits Malkus Eelco Rietsema
[email protected] [email protected] [email protected] 06 – 1321 4098 06 – 8333 9529 06 – 1004 2470
20
Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee (“DTTL”), its network of member firms, and their related entities.
DTTL and each of its member firms are legally separate and independent entities. DTTL (also referred to as “Deloitte Global”) does not provide services to clients. Please see
www.deloitte.nl/about for a more detailed description of DTTL and its member firms.
Deloitte provides audit, tax, consulting, and financial advisory services to public and private clients spanning multiple industries. With a globally connected network of member
firms in more than 150 countries and territories, Deloitte brings world-class capabilities and high-quality service to clients, delivering the insights they need to address their most
complex business challenges. Deloitte’s more than 200,000 professionals are committed to becoming the standard of excellence.
This communication contains general information only, and none of Deloitte Touche Tohmatsu Limited, its member firms, or their related entities (collectively, the “Deloitte
network”) is, by means of this communication, rendering professional advice or services. No entity in the Deloitte network shall be responsible for any loss whatsoever sustained
by any person who relies on this communication.
© 2014 Deloitte The Netherlands
