RISK MANAGEMENT

FIN 562 Risk Management

Professor Mark Shore

DePaul University

Week 9 Lecture Notes

[email protected]

Twitter @shorecap

Model Risk Management

Uses of Models • Credit decisions and credit exposures

• Liquidity management

• Derivatives valuation

• Calculation of VaR, Expected shortfall and other risk measures

• Asset management

• Fraud detection

• etc

Risk Management and Financial Institutions 5e, Chapter 25, Copyright © John C. Hull 2018

Models • Models approximate reality

➢Ideally try to keep the models simple

➢Assumptions are built into the models

• Models are considered part of operational risk

• Models need to be: • Documented

• Validated

• Used the way its intended

• Monitored

Regulatory Requirements

• April 2011 SR11-7 was published from the Fed to provide guidance to banks on model risk management

• Model risk: “ the potential for adverse consequences from decisions based on incorrect or misused model outputs and reports”

• A systematic approach to managing model risk

• Model risk can arise from ➢fundamental errors involving data, calculations, numerical

procedures, and assumptions

➢Inappropriate use of the model

➢Not understanding the limitations of a model

SR 11-7 continued

• SR11-7 contains directives on • Model development, testing, and documentation

• The role of model validation groups

• The use of vendor models

Risk Management and Financial Institutions 5e, Chapter 25, Copyright © John C. Hull 2018

Vendor Model

• Subject to the same validation as internal models

• Have a process to choose vendor models

• Does the vendor conduct ongoing monitoring of their models?

• Have information on data used to develop the model

• Contingency plan if the vendor stops supporting the model or goes out of business? ➢Vendor should make available proprietary data

Risk Management and Financial

Institutions 5e, Chapter 25,

Copyright © John C. Hull 2018

Finance vs Physics (page 572)

• The models of physics describe physical processes and are highly accurate. Their parameters do not change through time.

• The models of finance describe human behavior. They are at best approximations. Parameters do change through time

Developing & Testing Models

• Marking to Market ➢Valuation based on market prices

• Marking to Model ➢Valuation based on model pricing

➢Are assumptions built into the model?

• Stochastic Model ➢Model will have some randomness to the results; results can vary

• Deterministic Model ➢Model will have greater consistency (constant) of results

Marking the Prices of an Instrument to Market • Use prices quoted by market maker

• Use price at which financial institution has traded product

• Use prices from SEFs (Swap Execution Facilities) or interdealer brokers

• Use price indications

• Use model (marking to model)

Risk Management and Financial

Institutions 5e, Chapter 25,

Copyright © John C. Hull 2018

Risk Management and Financial

Institutions 5e, Chapter 25,

Copyright © John C. Hull 2018

Models for Less Actively Traded Products (page 578-579) In the case of less actively traded products, models play a key role in both pricing and hedging

➢ May have to call around for quotes on less liquid markets to create price market

What Makes A Successful Model 1) Ensures prices or results are consistent with market

prices or behavior

2) The model is a communication tool to understand the behavior of the market

3) The model is relatively simple to understand the behavior of the model and how it will perform in different environments

• Within-model hedging ➢Risk of changes in model variables assumed to be

uncertain (stochastic)

➢Example: Price changes in the underlying market

• Outside-model hedging ➢Risk of changes in model variables assumed to be

constant (deterministic)

➢Example: Black-Scholes model assumes volatility to be constant

Risk Management and Financial

Institutions 5e, Chapter 25,

Copyright © John C. Hull 2018

Model Building Missteps (page 581-582)

• Overfitting

• Overparametrization

• Making the model so complicated that the intended user will not be able to use it

• Black-Scholes has been successful because it is a simple interpolation tool that structures the trader’s thinking about the market

Model Due Diligence • Cost structure

➢Are costs built into your model?

➢Is there a cost when the model stops working?

• Back testing ➢Is it fitted to the data (curve fitting)?

• Forward testing ➢Out of sample testing

➢Does the model work in other environments

➢When does it work well and when does it fall apart

17

Portfolio Construction

Market selection

- Meets compliance standards

- Liquidity

- Market has characteristics of model

Risk allocation

- Model performance

- Cross correlation

Periodic review

- New markets

- Change in liquidity

- Enhancements to model

Model Due Diligence

• Black box: check for programming bugs

• Understand the data: How many years of data is used

• Curve fitting parameters

• Slippage

• Cost structure tested?

• Does the model search for patterns/ relationships or behaviors? ➢Is there causation to the correlation?

Correlations are not always causation “In statistics, a spurious relationship or spurious correlation[1][2] is a mathematical relationship in which two or more events or variables are not causally related to each other, yet it may be wrongly inferred that they are, due to either coincidence or the presence of a certain third, unseen factor (referred to as a "common response variable", "confounding factor", or "lurking variable").”

Source: https://en.wikipedia.org/wiki/Spurious_relationship

Here is a link to a few more spurious relationships

• http://www.tylervigen.com/spurious-correlations

Correlations are not always causation

• Be cautious of data mining!!

• Data mining is the process of finding anomalies, patterns and correlations within large data sets to predict outcomes.

• Using a broad range of techniques, you can use this information to increase revenues, cut costs, improve customer relationships, reduce risks and more.

https://www.sas.com/en_us/insights/analytics/data-mining.html

Model risk for Actively Traded Products • Little pricing risk in most cases

• Hedging performance can depend on the model used and should be carefully investigated

Risk Management and Financial Institutions 5e, Chapter 25, Copyright © John C. Hull 2018

Pricing Risk for Less Actively Traded Instruments • Use more than one model for pricing.

• Calibrate all models used to the prices of actively traded instruments

• Create reserve accounts so that profits are not recognized immediately

• Use weighted Monte Carlo to determine the range of models that price actively traded standard instruments consistently with the market.

Risk Management and Financial

Institutions 5e, Chapter 25, Copyright ©

John C. Hull 2018

Accounting

• FASB 157 and IASB 39 classify instruments as “held for sale” or “held to maturity”

• Those classified as held for sale have to be marked to market • Level 1: uses quoted prices in active markets

• Level 2: uses quoted prices for similar product in active markets or same product in non-active markets

• Level 3: requires valuation assumptions

Risk Management and Financial

Institutions 5e, Chapter 25,

Copyright © John C. Hull 2018

Thank You