Estimating Multiple Risk Measurements for Internal and External - - PowerPoint PPT Presentation

Estimating Multiple Risk Measurements for Internal and External Audiences Using a Manageable Approach: Stress Testing Structured Portfolios

YANIV GERSHON, STATE STREET JUAN CARLOS CALCAGNO, MOODY’S ANALYTICS

Outline

» Discuss project aimed to stress test State Street structured portfolio

• Describe standard modeling approach for diverse assets and regions
• Explain economic stress testing framework and assumptions
• Internal and external applications for

 Credit risk economic capital  Reserve requirements  Research, development, and investment

Overview of Project

CreditCycle – Behavioral Model Collateral Performance Economic Parameters (GDP, HPI, employment

Default Vector Prepayment Vector

CDR CPR

Cash-Flow Engine

Principal and interest payments Loss Estimation

Severity Vector

Severity

Choose your Macro Scenario ECAP OTTI Process Stress Test Program

Model Estimation Macro Forecast Assumptions

Deal-Level Behavioral Assumptions

Security-Level Cash-flows Different uses at State Street

DATA, MODELS AND ESTIMATION

5

Moody's Performance Data Service (PDS)

Our data and modeling team collects, normalizes and stores data

Data collection Scrubbing, validation & standardization Accessible Data Data warehousing

4.7199998 4.7199998 4.9333301 5.4699998 5.4699998 5.3666701 5.6399999 5.6399999 6.0666699 6.25 6.25 6.8666701 6.7122302 6.6847372 8.0736609 7.3008995 7.1676712 8.6520376 8.136611 7.7857828 9.1786413 8.6848841 7.9900608 9.4953976 9.2574759 8.0914431 9.7242556 9.1450596 7.4228692 9.7663956 8.7279348 6.5703249 9.6879368 8.2612009 5.9571881 9.386632 7.8519301 5.603909 8.9864655

Direct Download

Software Suite

00000000000000000 0000000000000000000000 00000000000000000000 0000000000000 000000000000000000000000 00000000000000000000000 0000000000000000000000 00000000000000000000 0000000000000 000000000000000000000000 00000000000000000000000 0000000000000000000000 00000000000000000000

Collect pool, deal, tranche and loan level data from originators, servicers and rating agencies. Over 13,000 deals back to 1980s and over 50 performance metrics 1 An automated program sorts and cleans the data, then a team of specialists manually validates and scrub key fields 2 Data warehouse is constructed from scrubbed data, and is made available for use in the software

• r accessible via direct download

3 Data sets, performance reports and indexes would be available to end users 4

Data Coverage ABS/RMBS Deals

6

Number of Unique Pools Market as Feb 2011 USA RMBS 11,689 Subprime 3,530 Alt-A 4,562 Jumbo 2,327 Other 507 Prime Conforming 1 Option Arms 762 USA HEL 275 HELOC 212 Home Equity/Closed End 63 Australia 136 Greece 9 Ireland 22 Italy 92 Netherlands 107 Portugal 30 Russia 12 Spain 226 UK 141 TOTAL 12,739 Asset Class Product Line Number of Unique Pools Market as Feb 2011 Auto 251 Prime 176 Marginal 17 Subprime 58 Motorcycles 49 Credit Cards 37 Bank 24 Charge 2 Retail 6 Student Loans 348 FFELP 238 Private 79 Mixed 31 TOTAL 685 Asset Class Product Line

Highlights of credit risk models

» A comprehensive approach based on a series of econometric models, each designed to capture a specific component of pool behaviour – pipeline, prepayment, default and severity » Model generates correlations between performance and macroeconomic data » Correlations are estimated using ALL active and inactive deals in the market » Time series includes data covering a full business cycle » Based on multiple regression analysis, deal specific vectors are generated under alternative economic scenarios » Standardized pool-level data and similar econometric framework is used for ALL ASSETS AND COUNTRIES

Econometric Model: Dynamic Panel Data

Pool performance time series

(e.g., CPR, CDR, LGD Vectors)

= f

Lifecycle component » Takes into account the shape of the performance curve over pools’ life Lifecycle component

Pool-specific quality component » Attributes (LTV, collateral type, region, etc) define quality across pools » Early delinquencies also serve as proxies for underlying pool quality » Economic conditions at origination matter for pool quality » Econometric technique accounts for other unobserved effects

Pool & Loan-level components Business cycle exposure component » Explicit connection between pool performance and macroeconomic drivers provides the ability to stress test holdings and run “what-if” scenarios Business cycle exposure component

Different subsets of macroeconomic variables are used for each vector and asset type

» Labor Market Indicators » Avg. Hourly Wages and Disposable Income » Net Worth and Debt Service Ratio » Interest Rate – Fed Funds STUDENT LOANS » Wages and Disposable Income » Employment ,Unemployment Rate-UI Claims » Real GDP » Interest Rate – Bank Prime Rate CREDIT CARDS » Home Prices » Existing Home Sales » Refinancing Activity » Unemployment Rate » Real GDP » Disposable Income and Wages » Fed Funds and Mortgage Composite Rates RMBS/HEL » Unemployment Rate-UI Claims » CPI New and Used Cars » Vehicles Sales & Car Registrations » Disposable Income » Real GDP » Personal Bankruptcy Rates » Interest Rate – Bank Prime Rate AUTO/MOTO LOANS & LEASES

Notes: All models include nonlinear terms for the lifecycle (age in months), and seasonality factors (month)

US Auto Loans, Leases, Motorcycle, RVs and Boats: Summary of model inputs

30 day delinquency 60 day delinquency 90+ day delinquency CDR Repossession Net Chargeoff CPR Principal Weighted Average Coupon (WAC) X Weighted Average Maturity (WAM) X Loan Type X X X X X X X X Unemployment Rate X X X X X X Prime Rate X Used Car Prices X X X X X X New Car Prices X X X GDP t, t-3 t, t-3 t, t-3 t, t-3 t, t-3 t, t-3 t, t-3 Unemployment Rate t t t t t t t Relative Unemployment Rate t t t t t t t Relative Prime Rate t Used Car Prices t t t t t t Relative Used Car Prices t t t t t Relative New Car Prices t t Automobile Sales t Automobile Registrations t t t t t t Personal Bankruptcies t t t t t t Mortgage Loan-to-Value Ratio t t t t t t 30 day delinquency t t 60 day delinquency t 90+ day delinquency t t Pipeline connections Group Variable Vector Origination Conditions Economic Conditions at Origination Current Economic Conditions

Example: USA Alt-A RMBS Deal (2006 vintage)

Constant Default Rate (annualized %) You can ask: “What would happen to my deal’s CDR under the 5 different scenarios given by Chief Economist Mark Zandi?”

Worst Case Scenario

USE OF MODELS AT STATE STREET

Overview

CreditCycle – Behavioral Model Collateral Performance Economic Parameters (GDP, HPI, employment

Default Vector Prepayment Vector

CDR CPR

Cash-Flow Engine

Principal and interest payments Loss Estimation

Severity Vector

Severity

Choose your Macro Scenario ECAP OTTI Process Stress Test Program

Model Estimation Macro Forecast Assumptions

Deal-Level Behavioral Assumptions

Security-Level Cash-flows Different uses at State Street

» Several functions within State Street are interested in analyzing Structured Assets performance

• Global Treasury – responsible for investment decisions and the Other Than

Temporary Impairment (OTTI) process

• Finance – responsible for State Street’s stress test program and the Fed

Comprehensive Capital Analysis and Review (CCAR) program

• Enterprise Risk Management - in its role of business units risk management

and Credit Risk Economic Capital

Overview

» State Street developed a process for a

firm wide macroeconomic scenarios that are used throughout the bank

• The scenarios are developed by a

macroeconomic forecasting firm with inputs from State Street’s Economics Team

• The scenarios are reviewed and approved

by a Scenario Review Board with members from different areas of the bank

• The scenarios are used for budgeting and

analysis purposes.

Internal macroeconomic scenarios

» Global Treasury is responsible for State Street’s Investment Portfolio (IP). » In this capacity, it makes decisions on purchase and sell of structured assets bonds » Global Treasury Risk, analyses the bonds on an ongoing basis and go through a

very robust OTTI process on a quarterly basis (as prescribed by SFAS 115-2).

Global Treasury

• Review of the full portfolio and

identification process for deep dive bonds

• Deep dive analysis on selected

bonds taking CreditCycle vectors as inputs

• Waterfall cashflows based on

CreditCycle vectors with analyst adjustments

» State Street Finance has been

designated to develop and deploy a stress test program for the bank.

» The Stress Test team conducts

tests on a quarterly basis. The test maybe on a macroeconomic level

• r specific to State Street either as

a whole or specific business units.

» State Street’s Stress Testing

Advisory Committee prioritizes the tests to be conducted

Finance

CreditCycles Models

Economic Capital

Capital Charge by Sub-Asset Type

Aug-10 Aug-11

» For credit Economic Capital (ECap) purposes State Street divides its portfolio into

two portions: Credit ECap for Securitized Assets (CESA) and Credit ECap for Non- Securitized Assets (CENSA)

» For CENSA we currently use a Multi-Factor Merton model approach

» For CESA we use the CreditCycles models

• Estimation is done on historical data
• Forecasts are calculated based on the Economic Capital

scenario approved by the Scenario Review Board

• Model outputs (CDR, CPR, and Severity) are then entered

into a cashflow engine (INTEX) for principal writedowns and interest shortfalls

• ECap is defined as Total lifetime Losses – OTTI
• Extremely helpful in identifying capital “Hogs” and manage

holdings accordingly

Sample results for the alternative scenarios

5 10 15 20 25 30

Constant Default Rate (%)

2 4 6 8 10 12 14 16 18 20

Constant Prepayment Rate (%)

10 20 30 40 50 60 70

Severity (%)

» Below are example of a US RMBS Prime deal “GSR Mortgage Loan Trust 2005-

6F”. Under the alternative scenarios the deal suffers different levels of losses.

• Under the Base case the deal does not suffer losses and therefore no OTTI charges.
• Under Stress Test and ECap the deal suffers a 1% and 3.4% loss respectively.

Base Stress ECap

» Same models used for different functions within the bank » Very transparent and intuitive » Easley used for loss allocation » Same econometric methodology across asset types increases speed of

implementation and makes it easier to explain to Senior Management

» Easley modified for special projects arise or when alternative scenarios are

considered

• European stress test
• Oil shock scenario
• Foreclosure crisis of 2010

Advantages of using the models