This document discusses risk management practices for hedge funds. It notes that risk management approaches differ across hedge funds based on their strategies. Common risk management tools discussed include Value at Risk (VaR), stress testing, concentration limits, drawdown management, and liquidity risk monitoring. The document cautions that no single risk management approach is best and that risk measures have limitations but can still provide useful insights if used appropriately.
Report
Share
Report
Share
1 of 57
More Related Content
Schachter_Hedge_Fund_Risk_Management_2015_09_17
1. Hedge Fund Risk
Management
Barry Schachter
Prepared for
Yale School of Management
Hedge Fund Program for Chinese Executives
September 17, 2015
2. The Varieties of Hedge Fund
Risk Management
Hedge funds are not all alike, and risk management is not the
same for all funds.
Hedge Funds Not all Alike
Risk Management Needs
Differ from HF to HF
No Single Approach to Risk
Management is Best
No external force (i.e.,
regulation) creates a
‘monoculture’ of Risk
Management
3. The Place of the CRO in the
Hedge Fund Organization
Most Hedge Funds Have
No CRO
Need for Independent Risk
Function Is a Flawed Idea –
Does Not Solve Agency or
Incentive Problems
Risk Management
Knowledge creation
Information exchange
Risk-Informed
Decisions
Many HF Risk Managers
only perform one or two
4. Knowledge Creation - Risk Prism
Good Risk Management Identifies Risks Like A Good
Prism Separates White Light
5. Risk Filter
Good Risk Management Mitigates Unwanted Risks
Like A Filter Eliminates Specific Colors from Light
6. Understanding Risk
Management
• From Harvard Business Review (March 2009)
o “Six Ways Companies Mismanage Risk” by René Stulz
• Mismeasurement of Known Risks
• Failure to take risks into account
• Failure in communicating the risks to top
management
• Failure in monitoring risks
• Failure in managing risks
• Failure to use appropriate risk metrics
7. Understanding Risk
Management
• These Six Are Better Thought of As the Recurring
Challenges of Risk Management
• A Successful Risk Management Function will Address
All Six
• But Cannot Eliminate Any of these Challenges
8. Assumptions Matter
• Assumptions Place Limitations on the Exploration of
Potential Risks
• It isn’t Enough to Question All Assumptions
• It is Necessary as Well To Identify Hidden
Assumptions
• It is Necessary to Consider Possibilities that are
Outside of Experience
• This Process Can Be Formalized as An Extension of
the Commonly Used “What Can Go Wrong?”
Analysis of Investment Ideas
10. Assumptions Matter
Global Financial Crisis Partly Blamed on MBS
Risk Models where HPA Rate Could Not Be
Negative for a Large, Geographically Diverse
Portfolio
…ratings agencies assigned their highest rating to
many mortgage-backed securities based partly on
the assumption that home prices would not fall.
Source: Investopedia (http://bit.ly/1icYojX)
…assumption that a diversified portfolio of US real
estate had little risk of falling in value.
Source: E. Rosengren. (http://bit.ly/1icYS9E)
…models used to price mortgage portfolios under-
weighted scenarios with large price declines.
Source: J. Kourlas. (http://bit.ly/1id1J2g)
11. Sources of Risk for HFs
Hedge
Fund
Economy
&
Markets
Investors
Data &
Systems
Trading
Counter-
parties
13. Trading Risk
Management
• Focused on Total Return, HFs Do Not Have a
Benchmark (other than cash)
• Active Risk Equals Total Risk
• Tracking Error is not a Meaningful Measure of
Portfolio Risk
• Risk Management of HF Trading Risk Does Not
Reduce to Solving A Standard Portfolio Optimization
Problem
• Aspects of the Optimization Problem Remain in
o Control of Common Factor Exposures
o Influence on Risk Allocation (Risk Budgets)
14. Trading Risk
Management
• Trading Risk Management Employs Multiple Risk
Measures to Support a Decision Process that is
Data-driven but Ultimately Qualitative
• In this Process Return Correlation is Not Sufficient
Information for Making Risk-Based Decisions
• Using Multiple Measures Adds to Confidence in
Decision-Making and Improving Communication
About Risk
• Using Multiple Measures Helps Capture Elements of
(Knightian) Uncertainty that Are Not Well
Represented by Correlation
15. Measures Used in Trading
Risk Management
• Value at Risk (VaR) (or Expected Shortfall)
• Stress Testing
• Sensitivities (e.g., Duration, PV01, OAS, Credit
Spread options)
• Notionals (e.g., Gross and Net Market Value)
• Beta-adjusted Exposures (single or multi-factor)
• P/L volatility
• Drawdown
• Measures of Liquidity
• Measures of Concentration
16. Value at Risk (VaR)
It Is the Magnitude of
Portfolio Loss that May
Be Exceeded
with a Specified
Probability (e.g., 5%)
or Confidence Level
This is a Statistical Risk
Measure (Derived from
Statistics of a Set of
Observations)
17. Calculating VaR
Many Ways to Calculate and Each
May Give Different Results, but All
Are Alternative Estimates of a
Distribution Function (Unknown, Assumed, or Fitted
by Backtesting)
General Approaches:
Parametric – Covariance Matrix
Historical – Empirical (Nonparametric)
Distribution
Monte Carlo – Covariance Matrix and
Simulation
Most Vendors Include Most of These Alternatives
18. Knowledge from VaR
Risk Aggregation – Allows
Combining Risks from Different
Sources
Risk Comparison – Creates A Common Measuring
Stick
Risk Allocation – Can Deduce Rules to Distribute
Risk (Based on Marginal Contribution to Portfolio
Risk)
Risk Attribution – Can Identify Exposures to
Components of Portfolio Risk
19. VaR Limitations
Assumes Historical Data Used Is
Appropriate for Forecast
Distribution Estimate
Many Subjective Elements in Modelling
Filling Missing Data, Implied Volatility
Modeling, Security Pricing Models, Proxy
Rules
False Precision & No Confidence Bands
VaR Calculation Methods Each Have
Unique drawbacks
Parametric: Normal, Nonlinear Positions
Historical: No ‘beyond the sample’
Monte Carlo: Parametric
20. VaR Problems – Imagined
Assumes Normal Distribution
Have Explained it is Not Necessary
Works Until You Most Need it
Need Risk Measures in All Environments,
Not Just Crises
Didn’t Predict the Crisis
Not Intended to Predict Returns
Excessive Reliance on VaR Contributed to the
Crisis
Anecdotally, I never saw much reliance
placed on VaR
21. Using Value at Risk
Limits
The Good
One of few truly risk-based measurements
Can be used on very complex portfolios
The Bad
Can be Gamed by the Risk Taker, and
May force selling in market decline and
Encourage buying in quiet market
Not Well Adapted to Some Strategies, like
Merger Arb, Distressed Credit
The Ugly
Has Been Accused of Contributing to
Systemic Risk (“Liquidity Black Hole”)
22. Black Swan Risk
• ‘Black Swan’ Made Familiar by Nassim Taleb
• Event that Cannot Be Anticipated, Has a Big
Impact, and is Explained only after the Fact
• Often Mischaracterized As Any Low Probability
Event or Emerging Risk
• Because A Black Swan Is Unanticipated, it Cannot
be Hedged (Beforehand)
• It is Claimed that Building General Resilience into
Risk Taking is the only Way to Manage This
23. Tail Risk Hedging
• ‘Black Swan’ Insurance
• Hedge against the Unknown
Unknowns
• Some Have Argued that All
Market Disruptions Exhibit
Increased Volatility and
Equity Sell-offs
• Tail Risk Therefore Is Usually A
Long Option Strategy
• Unlikely to be cost effective
if
o Options Fairly Priced (A
Cognitive Bias Issue)
o Measured over a Long Time
Interval (A Business Cycle)
24. Stressed VaR
Global Financial Crisis Partly Blamed on MBS Risk
Models where HPA Rate Could Not Be Negative for a
Large, Geographically Diverse Portfolio
…ratings agencies assigned their highest rating
to many mortgage-backed securities based partly
on the assumption that home prices would not
fall.
Source: Investopedia (http://bit.ly/1icYojX)
…assumption that a diversified portfolio of US
real estate had little risk of falling in value.
Source: E. Rosengren. (http://bit.ly/1icYS9E)
…models used to price mortgage portfolios
under-weighted scenarios with large price
declines.
Source: J. Kourlas. (http://bit.ly/1id1J2g)
25. Estimate Risk Using Volatile
Environment
Risk Changes Over Time Caused
Only by Portfolio Changes
(Look-back window never moves)
Stressed VaR
26. Reverse Stress Test
Scenario-based Stress Test:
What you do: Define price shocks to risk
factors, then revalue the portfolio
What you get: An estimate portfolio loss that
results from applying those shocks
Reverse Stress Test:
What you do: Define a portfolio loss level,
then investigate through what shocks that
may happen
What you get: Many alternative sets of price
shocks that all result in the defined loss level
28. Drawdown: Fall from a
Maximum
Drawdown in neither
inherently good nor
Bad
Volatility entails
drawdown
Expected
Drawdown
Increases as
Volatility Increases
Drawdown has a down
side, however
Drawdown
29. Return is about end points,
Drawdown is about the path
taken between end points
The path matters if it can affect
the outcome
Drawdown may reflect Bad
Portfolio Decisions
Drawdown may Trigger Investor
Redemptions
Why Manage Drawdown
30. Limit: At a certain magnitude of drawdown risk is
reduced or portfolio liquidated
Drawdown Limits Reduce Expected Return (In
General)
Intermediate Drawdown Thresholds may create
magnet effect (“pull to limit”)
Considerations
Portfolio Liquidity
Magnitude of Correlations (Multi-manager)
Trading Styles
Investor Liquidity Terms
Drawdown Limits
31. Concentration &
Diversification
• One or a Few Positions
Contribute
disproportionately to
Portfolio Risk
• Causes Too Much
Portfolio Sensitivity to
Idiosyncratic Events, or
Exposure to a Market
Factor
32. Constraints on Maximum Position Size
In Long/Short Equity, Commonly 10% of
AUM for Longs, 5% for Shorts
Constraints on Factor Exposures – In
Long/Short Equity Sometimes set at Zero
Exposure
Concentration &
Diversification
33. Trading Liquidity Risk
• The Risk to Realization of Gains on a Position
or Augmentation of Losses When Attempting
to Exit
• Liquidity is Defined In Terms of The Cost to
Exit
• Factors Affecting That Cost Are
o Position Size in Relation to Market Trading Activity
o The Transaction Costs of Exiting (Including Crossing
the Spread)
o The Impact of Speed of Exit on Price
34. Trading Liquidity Risk
• Optimizing Exit Strategy Is More Common
In Algorithmic Trading
• Monitoring Liquidity and Setting Liquidity
Limits Is The More General Approach
Used
35. Trading Liquidity Risk
• Monitoring Liquidity
o Exchange-traded – Position Size as Percent of
Average Daily Volume
o OTC – Two Dimensions
• The Product of the “Clip Size” and Number of Clips That
Can be Traded in a Day with Minimal Market Impact
• The stability of the ability to transact
• Liquidity Limits, If Used, are Strategy
Dependent
o May Take the Form, “No More than X% of Long (Short)
Exposure Requires More than 5 Days to Exit at 20% ADV”
36. Crowded Trade Risk
• A Special Case of Liquidity Risk
• Crowded Trade Origins
o Hedge Funds Follow Similar Strategies (E.g., Merger Arb)
o Hedge Fund Group Think (E.g., Global Macro often)
• Creates Latent Illiquidity, Manifests Only
when Hedge Funds as a Group Attempt to
Exit at the Same Time
• Difficult to Detect Beforehand, Except
Anecdotally
37. Crowded Trade Risk
Management
• Data Monitoring and Reporting
o Large Hedge Fund Ownership in Portfolio Names/Strategies
o Large Equity Short Interest
o Changes in correlations among assets
• In Some Cases, Maintaining Lower Leverage
(or More Unencumbered Cash)
38. Risk Parity - Crowded
Trade?
• Risk Parity Made
Popular By Bridgewater
and AQR
• Equities and Fixed
Income Balanced
weighted to Have
Equal Risk Contribution
(Requires Adding
Leverage to the FI
Share)
39. Risk Parity – Crowded Trade?
Problem:
bonds and
equities sell-off
at same time
1 August,
JPMorgan’s
index of risk
parity funds
lost 8.2 per
cent since the
beginning of
May;
Russell 3000
lost 9.2% over
same period
40. P&L Attribution
• Evaluate the Extent Gain/Loss Arises from
Unwanted or Unexpected Sources
• Use multiple approaches
o By strategy, or long vs. short Positions
o By single factor or multiple factor decomposition
41. Trading Risks from
Cognitive Biases
• Risk Taking Decisions are Subject to Bias
o https://en.wikipedia.org/wiki/List_of_cognitive_biases
• Important Biases for HF Risk Management
o Confirmation Bias
o Hindsight Bias
o Recency Bias
o Disposition Effect
o Round Number Bias
• Managed Through
o Data Analysis, e.g., holding period analysis
o Qualitative Discussion with Risk Takers, Adopting Heuristics
43. Quantification of
Counterparty Risk
• In General, Hedge Funds’ Approach is
Unsophisticated, Limited to Monitoring Current
Exposure
• Counterparty Exposure Measures
o Current Exposure
• Margin Posted (Required and Excess)
• OTC Swap and Option Exposure
o Potential Exposure – VaR-like Estimate
• Measures of Counterparty Default Risk
o Credit Spread and Credit Rating
o Stock Return and Implied Stock Volatility
44. Management of
Counterparty Risk
• Minimize Excess Margin
• Limit Rehypothecation (or Keep Margin/Collateral
at 3rd Party)
• Use Multiple Prime Brokers
45. Funding Liquidity –
Counterparty Margin
• Margin Calls Can Become An Existential Risk for a HF
o Insufficient Unencumbered Cash Requires forced Position
Liquidation
• HFs Target Significant Levels of Unencumbered
Cash
• Margin Should be Optimized to the Extent Feasible
o Prime Brokers Use Different Margining Approaches
o Allocate Trading Positions to Primes to Minimize total Margin
• Conduct margin portfolio stress tests
o Must Model Margin Agreement Terms
o Examine Impact of Changes in Market Environment and
Possible Changes in Required Margin
48. Operational Risks - Mitigation
• Trading
o Fat Fingers – Limit Trade Sizes
o Incorrect Orders – Internal Verification and Review
• Risk Measurement
o Bad Historical Prices – Integrity Checks
o Incorrect Position Set-up – Internal Verification for new
Positions
o Diagnostic Reports (after the fact)
• Proprietary Code
o Test Cases – User-performed Tests
o Cross-Validation
50. Investor Risk Disclosure
• Investors are better partners when they
understand the strategy and are Able to
Correctly interpret the risks and returns
• Disclosures are an Opportunity to Dispel
Misconceptions/Biases, E.g.,
o Interpretation of Maximum Drawdown
o Usefulness of Benchmark Tracking Errors
51. Investor Risk Disclosure
• Should Disclose Information that Allows Investors to
Determine Fund is Following Risk Guidelines in PPM
• Should Disclose Risk Decomposition Consistent with
Permitting Investors to Evaluate Strategy Drift
• I Would Not Participate In Risk Aggregation
Programs, such as Opera or Hedge Platform, as It is
Impossible to Stand Behind the Data After It has
Been Processed by a Third Party
• I Would Not Disclose Non-Public Regulatory
Information, such as form PF, as It May Be
Misleading as to the Fund’s Actual Risk
52. Investor Risks
Liquidity –
Redemptions
MFN” clauses
and side letters,
the death spiral)
Adverse affects on
strategy
implementation
Managing to
monthly P&L
Managing
drawdown
53. Emerging Risks
• Identification through Environment Scanning
o “…those who look only to the past or the present are
certain to miss the future.” John F. Kennedy (http://bit.ly/1IQWxXb)
• Known Unknowns that
o Cannot be Quantified (Knightian ‘uncertainty’),
o Are Not Priced,
o Nor Able to Be Hedged through Diversification
54. Emerging Risk
Measurement
• Big Data Analytics
o RecordedFuture.com – Threat Analysis
o DataMinr.com – Twitter as Information Discovery
o Cytora.com – Knowledge from Unstructured Public Data
• Brainstorming
Potential Scenarios
• Current Example
o El Nino
• Impact on Agriculture
• Transport
• Sensitivities
55. Emerging Risk - Example
• El Niño – Identified as a Likely to be “Very
Strong” (3rd time Since 1950)
• Impact Cited or Speculated
o Agriculture, E.g., California
o Transport, E.g. Transport
Cost from Asia to US
o Winter Resorts
• Estimate Price
Sensitivities (Severity)
• Estimate Likelihood
56. In Sum
• Have only Touched on Many Aspects of and Issues
in Hedge Fund Risk Management
• Hedge Fund Risk Management Defies A Simple
Description
• This is a good thing
• In Every Manifestation
Aimed at Improving
Risk-based Decisions
• Improving Performance,
not Avoiding Risk
57. Additional Readings
• “Parity Strategies and Maximum Diversification”
http://bit.ly/1JHmQQe
• “Extreme Risk Management” http://bit.ly/1JHmUzy
• “Financial Risk Management: A Practitioner’s Guide to
Managing Market and Credit Risk”, John Wiley & Sons, 2012
•
Editor's Notes
Single manager fund with outsources middle and back office and a few spreadsheets to multi-manager, multi-strategy, multi-asset class funds, with hundreds of employees and internally run middle and back offices.
Strategies differ dramatically as well, encompasing many varieties from discretionary market-neutral, long short equity to classic directional global macro
Additionally, risk management as it is practiced in hedge funds is influenced by idiosyncrasies of organizational culture, innovation and individualism.
I will talk about Common themes in Risk Management in HFs.
In such cases the risk management function is the responsibility of senior front office management. Where there is an individual with the CRO title, that person should report to the President/CEO or if those roles do not exist to the CIO. The organizational structure should facilitate engagement with management and risk takers in risk discussions of all kinds.
In truth, risk management is everyone’s job, and the HF culture should emphasize that.
You can formalize the process of questioning assumptions
You can formalize the process of questioning assumptions
IC: Information Coefficient; IR: Information Ratio (alpha return/std dev of tracking error)
They are communication tools. Volatility and correlation isn’t enough to understand the risk in a portfolio. Partly because that doesn’t capture the “uncertainty” that isn’t risk and partly because that misrepresents the risk that is embedded in a portfolio.
They are communication tools. Volatility and correlation isn’t enough to understand the risk in a portfolio. Partly because that doesn’t capture the “uncertainty” that isn’t risk and partly because that misrepresents the risk that is embedded in a portfolio.
It is not bad or good in itself.
Useful for exploring relationships in a portfolio
Standard normal distribution 3 standard deviations event is is about a 99.9% (99.87) event, or a 1 in 1,000 event. For a t-distribution with 3 degrees of freedom, a 99.9% event is about a 10 standard deviation event. That is, when someone says, the market move today was 3 standard deviations, that meant it is a modestly infrequent event if stock returns are distributed normal, but not an infrequent event at all if stock returns are distributed t(3), about 1 in 20 event.
Standard normal distribution 3 standard deviations event is is about a 99.9% (99.87) event, or a 1 in 1,000 event. For a t-distribution with 3 degrees of freedom, a 99.9% event is about a 10 standard deviation event. That is, when someone says, the market move today was 3 standard deviations, that meant it is a modestly infrequent event if stock returns are distributed normal, but not an infrequent event at all if stock returns are distributed t(3), about 1 in 20 event.
It is not bad or good in itself.
Useful for exploring relationships in a portfolio
Building resilience is somewhat vague.
Universa
You can formalize the process of questioning assumptions
You can formalize the process of questioning assumptions
The Risk is to the Realization of Gains on a Trade or the Augmentation of Losses When Attempting to Exit A position
The Risk is to the Realization of Gains on a Trade or the Augmentation of Losses When Attempting to Exit A position
It is not bad or good in itself.
Useful for exploring relationships in a portfolio
It is not bad or good in itself.
Useful for exploring relationships in a portfolio
It is not bad or good in itself.
Useful for exploring relationships in a portfolio
It is not bad or good in itself.
Useful for exploring relationships in a portfolio
It is not bad or good in itself.
Useful for exploring relationships in a portfolio
Practically speaking, many of these risks will be caught after the fact. When that is the case, the goal is speed of correction.
It is not bad or good in itself.
Useful for exploring relationships in a portfolio
It is not bad or good in itself.
Useful for exploring relationships in a portfolio
It is not bad or good in itself.
Useful for exploring relationships in a portfolio