financial risk management prof. jeff (yuqing) shen, uc berkeley section 4: risk management for hedge...
Post on 20-Dec-2015
217 views
TRANSCRIPT
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Section 4: Risk Management for Hedge Funds
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
2
The BIG Picture
• Theory
•Efficient Frontier
•Asset/Liability Framework
•Alpha/Beta risk budgeting
• Practice
•Pension Crisis
•GM Pension Plan
•Yale Endowment
• Theory
•VaR
•Stochastic behavior of asset
returns
•Worst case scenario
• Practice
•Investment bank VaR
•Bear Stearns
• The Theory
•Active Portfolio Management
•Quantitative equity models
•Hedge fund investment
• The Practice
•Hedge fund industry
•Aug 2007 crisis for quant
Investment BankInstitutional Investor Hedge Fund
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Active risk
3
Exp
ecte
d a
lph
a %
Risk-controlled Active Traditional Active
The Risk Budget — How much risk?
Index Fund
Active risk %
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Asset mix with in US equity
4
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Active management trend in international equity
5
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Active management in US bond
6
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Lots of managers
7
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
8
The BIG Picture
• Theory
•Efficient Frontier
•Asset/Liability Framework
•Alpha/Beta risk budgeting
• Practice
•Pension Crisis
•GM Pension Plan
•Yale Endowment
• Theory
•VaR
•Stochastic behavior of asset
returns
•Worst case scenario
• Practice
•Investment bank VaR
•Bear Stearns
• The Theory
•Active Portfolio Management
•Quantitative equity models
•Hedge fund investment
• The Practice
•Hedge fund industry
•Aug 2007 crisis for quant
Investment BankInstitutional Investor Hedge Fund
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
9
Active return: alpha
Benchmark portfolio of stock and bond with weights wstock and wbond, where wbond = 1 - wstock
The tactical bets are betstock and betbond, with betstock = -betbond.
Return for benchmark = wstock*Rstock + wbond*Rbond
Return for tactical allocation portfolio = (wstock + betstock) *Rstock + (wbond + betbond) *Rbond
Alpha = Return for tactical allocation portfolio - Return for benchmark
=betstock (Rstock - Rbond)
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
10
Historical Tracking error
Historical Tracking error calculation:
2
1 1
1
1
1
T
t
T
tttt TT
TE
Tracking error annualization:
tTETE year ain periods ofnumber
Issues:– Assume time independence of alpha
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
11
Information Ratio
Information ratio calculation:
Information ratio annualization:
Error Tracking
AlphaRation Informatio
Error Tracking dAnnualilze
Alpha dAnnualilzeRation Informatio dAnnualilze
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
12
Typical IR in the industry
Percentile IR
90 1.00
75 0.50
50 -
25 (0.50)
10 (1.00)
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
13
The fundamental law of active management
IR: Information Ratio
IC: Information Coefficient; skill of return forecasting; correlation (forecast, realization)
BR: Breadth; number of independent bets per year
IR IC B R *
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
14
Example: What’s expected IR?
Case 1: Stock picker, 200 stocks, IC=0.02, quarterly rebalancing
Case 2: Market Timer, IC=0.10, 4 bets per year
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
15
Example: What’s expected IR?
Case 1: Stock picker, 200 stocks, IC=0.02, quarterly rebalancing
Case 2: Market Timer, IC=0.10, 4 bets per year
IC = 0 .0 2 , B R = 2 0 0 * 4 = 8 0 0
IR = 0 .0 2 * 8 0 0 0 5 7.
IC = 0 .1 0 , B R = 4
IR = 0 .1 0 * 4 0 2.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
16
Implications of the fundamental law of active management
Be good
Given some skill, bet as often as possible
Diversify bets
Don’t market time
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
17
It’s hard to distinguish skill from luck
We have seen a top quartile manager has an IR of 0.5. How long must we observe such a manager to measure the IR with 95% confidence?
We want the t-statistics, the ratio of the IR to its standard error, to exceed 2
But requiring so long a time series is quite problematic. Things change!
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
18
The BIG Picture
• Theory
•Efficient Frontier
•Asset/Liability Framework
•Alpha/Beta risk budgeting
• Practice
•Pension Crisis
•GM Pension Plan
•Yale Endowment
• Theory
•VaR
•Stochastic behavior of asset
returns
•Worst case scenario
• Practice
•Investment bank VaR
•Bear Stearns
• The Theory
•Active Portfolio Management
•Quantitative equity models
•Hedge fund investment
• The Practice
•Hedge fund industry
•Aug 2007 crisis for quant
Investment BankInstitutional Investor Hedge Fund
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Defining Risk as Standard Deviation
That’s what Harry Markowitz used.
It has several important and useful properties:– Symmetric– Well-understood statistical properties.– Machinery exists for aggregation from asset to portfolio.
– Predictable
2 TP P P h V h
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Issues
Magellan FundJ anuary 1973 - September 1994
0.00%
5.00%
10.00%
15.00%
20.00%
25.00%
30.00%
35.00%
40.00%
-25% -20% -15% -10% -5% 0% 5% 10% 15% 20%
Return
Non-normal distributions– Fat tails (kurtosis)– Skewness
Other choices:– Semivariance– Shortfall probability– Value at Risk
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Why Risk Models?
Let’s look at how we aggregate portfolio risk:
For N assets, we require N(N+1)/2 parameters. If N=1,000, we need to estimate 500,500 parameters.
That is the challenge.
2
2 2 2 2 2 21 1 2 2
1 2 12 1 3 13 1 1,2 2 2
TP P
N N
N N N N
h h h
h h h h h h
h V h
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Problems with historical risk
Let’s think about the number of parameters.
We need to estimate N(N+1)/2 parameters, and we have NT observations. We require at a minimum, 2 observations per parameter estimate. (How can we estimate a variance from 1 number?)
Hence, NTN(N+1), or T>N. This causes problems when we are looking at monthly returns for 1,000 assets.– Technical version: unless T>N, we will estimate a singular covariance
matrix. What does that mean, mathematically? Intuitively?
And even if we had 1,000 months of data (or more), we know that assets and markets change over time. We also regularly observe new assets.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Single Factor Model Sharpe’s Market Model
– This predated CAPM. It isn’t the same thing.
Decompose returns into market and residual components.
Postulate that residual components are uncorrelated.
2
21
22
2
0 0
0
0
mkt
Tmkt
N
r
r β θ
V β β Ω
Ω
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Sharpe’s Market Model
Solves the number of parameters problem:– For N assets, it requires 2N+1 parameters.– N betas, N residual risks, 1 market volatility.
Problem with this model: it doesn’t capture observed correlation structure in the market. Are Exxon and Chevron correlated only through their market exposure?
Advantage: simplicity makes this useful for back-of-the-envelope calculations.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Factor Models of Risk
These are extensions of Sharpe’s approach, designed to capture real market issues.
Separate returns into common factor and specific (idiosyncratic) pieces. We choose K factors, where K<<N. The covariance matrix is then:
r X b u
T V X F X Δ
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Choosing the Factors
Art not science.
Three general approaches:– Fundamental factors (BARRA, Northfield)
Industries and investment themes.– Macroeconomic factors (Salomon RAM, BIRR model)
Industrial productivity, inflation, interest rates, oil prices…– Statistical factors (Quantal, Northfield)
Use statistical factor modeling, principal components analysis…
These three approaches involve different estimation issues, and exhibit different levels of effectiveness.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Fundamental Models
Typically about 60 factors for a major equity market.
Calculate factor exposures, X, from fundamental data.– Industry membership.– Risk index exposures (e.g. value based on B/P)
Run monthly cross-sectional GLS regressions to estimate factor returns.
Use N observations to estimate K factor returns.
11 1 T Tb X Δ X X Δ r
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Aside: Factor Portfolios
Our extensive use of fundamental models makes it worth understanding factor portfolios in more detail.
We estimate factor returns as:
This equation has the form:
Each estimated factor return, bj, is a weighted sum of asset returns. We can interpret those weights as the asset weights in a factor portfolio, or factor-mimicking portfolio.
11 1 T Tb X Δ X X Δ r
T b H r
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Factor Portfolios
The columns of H contain the portfolio weights, with one column for each factor.
The GLS estimation approach guarantees that factor portfolio-j has:– Unit exposure to factor-j– Zero exposure to all other factors.– Minimum risk.
Industry factor portfolios are typically fully invested, with long and short positions.
Risk index factor portfolios are typically net zero investments, with positions in every stock.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Macroeconomic Models
Typically about 9 factors for a major equity market.
Calculate the change (or shock) in each macrovariable each month.
Estimate exposure to such shocks, stock by stock, using time-series data.
This approach requires NK parameter estimates.
1
K
n k nk nk
r t b t X t
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Statistical Models
Start with only returns data.
Use statistical analysis to determine number and identity of most important factors.– While this approach sounds completely objective, it involves many subjective decisions, e.g.
choosing portfolios to build an initial historical covariance matrix.
This approach implicitly assumes that factor exposures are constant over estimation period.
Factors change from month to month.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Performance and Uses* Fundamental models
– In general, best risk forecasting out-of-sample, but dependent on choosing the right factors.
– Intuitive factors also useful for performance attribution and alpha forecasting.
Macroeconomic models:– Poor at risk forecasting.– Direct macroeconomic connections can be useful for alpha forecasting.
Statistical models:– Best in-sample forecasts. Will outperform fundamental models with poorly chosen
factors.– Misses factors whose exposures change over time (especially momentum).– Not useful for performance attribution. Difficult to use for alpha forecasting.
*See Gregory Connor, “The Three Types of Factor Models: A Comparison of their Explanatory Power.” Financial Analysts Journal, May-June 1995, pp. 42-46.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Covariance Matrix Estimation
For fundamental and macroeconomic models (and even to some extent for statistical models), we still need to estimate the covariance matrix, given the factor return history.
We also need to estimate the specific (idiosyncratic) risk matrix.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Example: BARRA Approach*
Want best forecast of portfolio risk over the next several months. (Horizon will depend on use.)
Given that risk varies over time, we would like to overweight more recent observations.
At the same time, we have the parameter estimation challenge: we want T>>K. Lowering the weight on historical observations effectively lowers T.
BARRA US Equity approach combines exponential weighting, GARCH modeling of market volatility, and a special specific risk model.
*This does not exactly describe current BARRA approach.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
BARRA US Covariance Matrix
Historical monthly data back to 1973. Between 65 and 70 factors.
Step 1: Use exponential smoothing to estimate factor covariance matrix:
Step 2: Build a GARCH model for market volatility.
1
1
1ˆ 1
1
T
i i j jt
ij T
t
r t r r t r Exp T tT
T Exp T t
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Market Volatility Model
Apply GARCH model to market volatility (i.e. 1 factor):
2
2 2 2
, ~ 0,
1 1
mktr t t t N t
t t t
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
GARCH Model Intuition
We can rewrite this model by defining:
Then the GARCH model becomes:
In this format, we can see that is the long-term (unconditional) volatility, measures mean reversion in volatility, and measures how current shocks impact volatility.
2,1
2 2 2 2 2 21 1 1t t t t
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Step 3: Specific Risk Model
Our monthly factor return estimations also estimate specific returns:
We could calculate historical specific risk. Instead we estimate the following model:
We build models for S and v.– What are the estimation issues around this?
r X b u
2
2
1
1
1
n n
N
nn
u t S t v t
S t u tN
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Step 4: Combining these Steps
The challenge is to adjust the factor and/or specific risk matrices so that we can match the GARCH market volatility forecast.
BARRA does this by rescaling the systematic component of the factor covariance matrix.
2 T T Tmkt mkt mkt mkt mkt mkt mkts h V h x F x h Δ h
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Testing Risk Forecasts
Given r(t) and (t), how can we test whether (t) is a good risk forecast?
Bias test:– Convert returns to standardized outcomes:
– The bias statistic is the sample standard deviation of these outcomes:
– If the bias>1, we have under-predicted risk, and vice versa.
r tx t
t
| 1,bias StDev x t t T
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Testing Risk Forecasts
Statistical significance: Remember that for normally distributed random numbers:
The bias test estimates whether we are accurate on average. We can apply it to total, residual, common factor, and specific risk.
We can use further tests to see if our forecasts are above average when realized risk is above average, etc.
2
SET
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Total and Active Risk
Given the covariance matrix, we can estimate total and active risk:
We can also estimate the correlation of returns from two portfolios:
2
2
T T TP P P P P P P
T T TP PA PA PA PA PA PA
h V h x F x h Δ h
h V h x F x h Δ h
,TA B
A BA B
Corr r r
h V h
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
What could go wrong with the equity risk models?
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
44
The BIG Picture
• Theory
•Efficient Frontier
•Asset/Liability Framework
•Alpha/Beta risk budgeting
• Practice
•Pension Crisis
•GM Pension Plan
•Yale Endowment
• Theory
•VaR
•Stochastic behavior of asset
returns
•Worst case scenario
• Practice
•Investment bank VaR
•Bear Stearns
• The Theory
•Active Portfolio Management
•Quantitative equity models
•Hedge fund investment
• The Practice
•Hedge fund industry
•Aug 2007 crisis for quant
Investment BankInstitutional Investor Hedge Fund
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
45
Hedge Fund
Introduction
Demand
Supply
Issues
Outlook
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Hedge Fund Industry Questions
What are the pros and cons of hedge fund investment style?
What’s the future outlook for the hedge fund industry given the demand and supply landscape?
46
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
47
What is a Hedge Fund?
The term “hedge fund” is not defined or used in the federal securities laws
– Today the term refers to a variety of pooled investment vehicles loosely regulated and not registered under federal securities laws as public corporations, investment companies, or broker-dealers
– Hedge funds differ substantially in their investment objectives, use of different financial instruments, exposure to various markets and risk and return objectives
– Other generally accepted attributes include: ability to use leverage ability to use short sales to mitigate or increase risk performance based compensation schedule
Relative ValueRelative Value
Hedge Fund Strategy Classifications
Opportunistic/Opportunistic/MacroMacro
Merger Arbitrage/Merger Arbitrage/Event DrivenEvent Driven
Long/Short EquitiesLong/Short Equities Distressed SecuritiesDistressed Securities Short SellingShort Selling
47
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
48
Hedge fund industry AUM
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
49
Performance
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
50
The Advantage of Long-Short Investing
Long-short approach gives hedge funds a greater opportunity set to benefit from alpha
In most scenarios, long-short managers are able to reduce risk and increase expected return potential, relative to long-only managers
Efficient Frontier Scenario 1: Efficient Frontier Scenario 1: SPX, NDX Returns > 0SPX, NDX Returns > 0
-10
0
10
20
30
40
50
0 20 40 60 80Expected Risk
Exp
ect
ed
Re
turn
Long-Short Port Fully Invested1 Long Only Portfolio
Note: Borrowing and lending costs are assumed to be zero in these examples.1. Assumes the long-short portfolio is fully invested at all points, with leverage only to the degree needed to remain fully invested.Source: Morgan Stanley Quantitative Strategies.
Efficient Frontier Scenario 2: Efficient Frontier Scenario 2: SPX, NDX Returns < 0SPX, NDX Returns < 0
-50
-40
-30
-20
-10
0
10
0 20 40 60 80 100Expected Risk
Exp
ect
ed
Ret
urn
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
51
Other Advantages
Unconstrained by benchmark tracking considerations
Fewer regulatory hurdles
Organizational focus
Compensation similar to holding a real cal option
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
52
Hedge Fund
Introduction
Demand
Supply
Issues
Outlook
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
53
Demand to increase
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Demand profile of hedge funds
54
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Institutional demand for hedge fund
55
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Secular trends on AUM
56
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
57
Hedge Fund
Introduction
Demand
Supply
Issues
Outlook
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
58
Supply side dynamics
Low barrier to entry ($30-50K in legal, accounting and audit expense to start)
High incentive due to significant upside potential
Prime-brokers and fund-of-funds facilitate the distribution channel
Large institutional managers venture into hedge fund space as well
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
59
Fee structure
0.00%
1.00%
2.00%
3.00%
4.00%
5.00%
6.00%
7.00%
-10% -5% 0% 5% 10% 15% 20%
Return of fund
Fee
Note: Assume 2% base management fee, 20% performance fee and 0% hurdle rate
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Hedge fund return dispersion
60
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
New landscape of hedge funds
61
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Alignment
62
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
63
Hedge Fund
Introduction
Demand
Supply
Issues
Outlook
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
64
Persistence
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
65
Transparency
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
66
Scale and performance
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
67
Is hedge fund truly alpha or beta?
Source: Bridgewater Associates
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
68
A worrying picture
Source: Bridgewater Associates
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
69
Doing the same thing?
Source: Bridgewater Associates
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
70
Hedge Fund
Introduction
Demand
Supply
Issues
Outlook
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
71
Outlook #1: the squeeze on traditional long-only management
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
72
Outlook #2: Institutional quality hedge fund is the future
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
73
Outlook #3: Pure alpha is the future
Alpha correlation to
– US Equity
– International equity
– Fixed income
– Commodity
– Other major Beta risk premium
…should be ZERO
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
74
The BIG Picture
• Theory
•Efficient Frontier
•Asset/Liability Framework
•Alpha/Beta risk budgeting
• Practice
•Pension Crisis
•GM Pension Plan
•Yale Endowment
• Theory
•VaR
•Stochastic behavior of asset
returns
•Worst case scenario
• Practice
•Investment bank VaR
•Bear Stearns
• The Theory
•Active Portfolio Management
•Quantitative equity models
•Hedge fund investment
• The Practice
•Hedge fund industry
•Aug 2007 crisis for quant
Investment BankInstitutional Investor Hedge Fund
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Quant equity models
1. Value and Momentum model
2. Earning Surprise model
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley76
1. Value & Momentum Screen
Relative value and momentum equally weighted to calculate composite rank
Price Momentum
Earnings Momentum
Composite Rank
Relative Value (50%)
PriceAcceleration
orP/CF Model P/S Model
Momentum (50%)
Source: Smith Barney.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley77
Value & Momentum Model
Cumulative Performance of the Cumulative Performance of the Top-Over-Bottom DecilesTop-Over-Bottom Deciles
Monthly Return Spread from the Monthly Return Spread from the Top-Over-Bottom DecilesTop-Over-Bottom Deciles
Source: Smith Barney. Source: Smith Barney.
0
1
2
3
4
5
6
7
Most attractive
Least attractive
Universe
Launch of the Model
Modify Model due to IFRS
-12
-8
-4
0
4
8
12
16
20
%
Monthly Return SpreadSix Month Moving Average
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley78
STOCK RANKING
Company Name
Sector
Price (€)
Composite Rank
Relative Value Rank
Momentum Rank
SB Analyst Rating
IBES Mean Rec.
This Month’s Rank
Last Month’s Rank
ING GROEP Diversified Financials 22.93 1 1 1 1M 2.1 1 1 FORTIS Diversified Financials 21.80 1 1 2 2M 2.8 1 1 SUEZ Utilities 20.72 1 3 1 2H 2.4 1 2 ASSICURAZIONI GENERALI Insurance 24.79 1 1 3 2M 2.8 1 1 AXA Insurance 20.18 1 1 3 1M 2.3 1 1 BBVA Banks 12.57 1 2 2 2M 2.8 1 4 HBOS GROUP Banks 11.95 1 2 3 1M 2.2 1 2 SWISS REINSURANCE CO Insurance 54.04 1 1 4 2H 2.1 1 4 CREDIT SUISSE Diversified Financials 32.91 1 3 2 2H 2.4 1 3 BASF Materials 55.01 1 5 1 2.1 1 3 BARCLAYS Banks 7.99 1 1 4 3M 3.1 1 1 ALLIANZ Insurance 97.00 2 2 4 2H 2.1 2 3 BNP PARIBAS Banks 54.25 2 2 4 1M 2.1 2 2 ABN AMRO HOLDING Banks 19.14 2 2 5 2M 2.8 2 2 AVIVA Insurance 9.16 2 1 6 2M 2.7 2 1 SHELL T & T Energy 6.99 2 5 2 2L 3.0 2 2 TESCO Food & Staples Retailing 4.60 2 5 2 1M 2.2 2 3 LLOYDS TSB GROUP Banks 6.87 2 6 2 2M 3.2 2 3 DEUTSCHE BANK NAMEN Diversified Financials 66.84 3 1 6 2.3 3 3 BSCH BCO SANTANDER CENTR
Banks 9.26 3 3 6 2M 2.7 3 2 UBS NAMEN Diversified Financials 64.98 3 4 4 1M 1.8 3 4 DEUTSCHE TELEKOM Telecommunication
Services 15.37 3 4 5 1M 2.0 3 5
ENI Energy 20.34 4 5 4 2M 2.4 4 4 ROYAL BANK OF SCOTLAND Banks 24.27 4 4 5 1M 2.0 4 9 SOCIETE GENERALE Banks 79.35 4 7 2 1M 2.4 4 1 ROYAL DUTCH PETROLEUM CO
Energy 46.70 4 5 5 2L 2.8 4 4 BT GROUP Telecom Services 3.02 5 2 8 1M 3.4 5 3 E. ON Utilities 66.21 5 5 5 1M 2.0 5 4 NOKIA CORP Tech Hardware & Equip 11.81 5 7 4 3H 2.6 5 6 DAIMLERCHRYSLER Auto & Components 34.19 5 1 10 3.0 5 5 TELEFONICA Telecom Services 13.25 6 7 4 1M 2.2 6 6 CARREFOUR Food & Staples Retailing 40.66 6 3 8 3M 2.7 6 5 ANGLO AMERICAN (GB) Materials 18.42 7 7 6 2H 3.2 7 9 PHILIPS ELECTRS (KON.) Cons Durables & Apparel 20.53 7 6 6 2.1 7 4 TOTAL Energy 182.70 7 6 7 1L 1.9 7 9 BP Energy 8.17 7 7 6 1L 2.2 7 9 GLAXOSMITHKLINE Pharma & Biotechnology 17.53 8 6 8 2L 2.6 8 6 TELECOM ITALIA ORD Telecom Services 2.91 8 6 8 1M 2.6 8 8 VODAFONE GROUP Telecom Services 2.05 9 8 7 3H 2.1 9 6 ASTRAZENECA Pharma & Biotechnology 30.38 9 5 9 2M 2.8 9 8 NESTLE Food Beverage & Tobacco 208.60 9 6 9 1M 2.2 9 10 HSBC HOLDINGS (GB) Banks 12.19 9 7 8 3M 3.2 9 1 NOVARTIS Pharma & Biotechnology 35.61 9 7 9 2L 2.3 9 9 DIAGEO Food Beverage & Tobacco 11.06 9 7 9 2L 2.9 9 8 ROCHE HOLDING GENUSS Pharma & Biotechnology 82.60 9 6 10 1M 2.0 9 9 SIEMENS Capital Goods 60.77 10 8 8 2.3 10 9 ERICSSON (LM) B Tech Hardware & Equip 2.17 10 6 10 2H 2.6 10 9 LOREAL Household & Personal
Products 60.20 10 10 8 2M 2.8 10 10
UNILEVER NV CERT Food Beverage & Tobacco 52.00 10 9 10 1M 3.2 10 10 SAP STAMM Software & Services 122.61 10 10 10 2H 2.3 10 10 The prices in the table above are taken from the close of 4 April 2005. Sources: STOXX, IBES, Factset and Smith Barney.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley79
2. Earnings Surprise Model
Analysts on average tend to overestimate earnings
Average Analysts’ Forecast Errors Average Analysts’ Forecast Errors (1990 – 03) (1990 – 03)
Average Analysts’ Forecast Errors for One to 11 Average Analysts’ Forecast Errors for One to 11 months Prior to Announcement (1990 – 03)months Prior to Announcement (1990 – 03)
-20%
-15%
-10%
-5%
0%
5%
1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
Analysts Pessimistic
Analysts Optimistic
-40%
-35%
-30%
-25%
-20%
-15%
-10%
-5%
0%
-11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1
Large-Caps Small-Caps
Sources: IBES, MSCI and Smith Barney. Sources: IBES, MSCI and Smith Barney.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley80
Price Impact of Earning Surprises
We find an asymmetric effect of surprises to share prices
The effect is consistent across time but seems more pronounced in periods of slower economic growth
Average Market Relative Return Around the Earnings Announcement Period for SUE Quintiles, 1990-04
-10%
-8%
-6%
-4%
-2%
0%
2%
4%
6%
-65 -55 -45 -35 -25 -15 -5 5 15 25 35 45 55 65
Days Around Earnings Announcement
Quintile 5(PositiveSurprises)
Quintile 4
Quintile 3
Quintile 2
Quintile 1(NegativeSurprises)
Sources: IBES, MSCI and Smith Barney.
Cumulative Market Relative Performance for Positive and Negative Surprises 65-Days Around the Announcement Period
-20
-15
-10
-5
0
5
10
15
199019911992199319941995199619971998199920002001200220032004
Negative Surprises Positive Surprises Sources: IBES, MSCI and Smith Barney.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley81
Can We Predict Earnings Surprises?
Likely Drivers of Earnings SurprisesLikely Drivers of Earnings Surprises
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley 82
Earnings Surprise Model
Estimated Factor Sensitivities (1990 – 03)Estimated Factor Sensitivities (1990 – 03)
Univariate OLS Model Multivariate OLS Model Logit Regression Model
(Intercept) -1.26* -2.45*Return on Equity 0.21* 0.14* 0.25*Earnings Revisions FY1 0.16* 0.12* 0.25*Price Momentum (1-year t stat) 0.16* 0.10* 0.18*Market Capitalisation 0.07* 0.02* 0.05*Previous Surprise 0.34* 0.21* 0.43*Direction of Next EPS Forecast 0.63* 0.14* 0.24*
Source: Smith Barney.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley 83
Predictive Accuracy of theEarnings Surprise Model We observe a monotonic increase in the actual SUE score as we We observe a monotonic increase in the actual SUE score as we
move from negative to positive predicted SUE quintilemove from negative to positive predicted SUE quintile
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
Large Negative Small Negative No Surprise Small Positive Large Positive
Predicted Surprise Quintile
Aver
age
Real
ised
SUE
Sco
re
Source: Smith Barney.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley84
Earnings Surprise Model
Pan-European Earnings Surprise ForecastsPan-European Earnings Surprise Forecasts
Pan-European Earnings Surprise Forecasts
0
20
40
60
80
100
120
140
Large Negative Small Negative No Surprise Small Positive Large Positive
0.0%
2.0%
4.0%
6.0%
8.0%
10.0%
12.0%
14.0%
16.0%
18.0%
No. of Stocks (LHS) % of Total Market Cap (RHS)
Source: Smith Barney.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley85
Earnings Surprise Model
Earnings Surprise Forecasts by Industry
Number of Stocks Reporting
% of Sector’s Market Cap
Number of Large Positive SUE
Number of Small Positive SUE
Number In Line
Number of Small Negative SUE
Number of Large Negative SUE
Automobiles & Components 6 46.0% 0 2 0 4 0 Banks 29 16.6% 14 4 4 5 2 Capital Goods 29 39.0% 16 5 6 1 1 Commercial Services & Supplies 4 5.9% 1 0 1 2 0 Consumer Durables & Apparel 16 45.2% 6 3 5 1 1 Diversified Financials 12 17.6% 3 1 3 2 3 Energy 14 60.6% 7 3 3 1 0 Food & Staples Retailing 7 54.7% 4 1 0 1 1 Food Beverage & Tobacco 7 10.6% 4 2 0 1 0 Health Care Equipment & Services 6 15.1% 4 1 0 1 0 Hotels Restaurants & Leisure 5 9.6% 4 0 1 0 0 Household & Personal Products 3 35.5% 3 0 0 0 0 Insurance 13 16.1% 3 4 2 3 1 Materials 21 23.4% 9 0 5 3 4 Media 17 10.3% 4 2 5 2 4 Pharmaceuticals & Biotechnology 13 39.3% 3 2 3 3 2 Real Estate 12 19.9% 5 3 2 1 1 Retailing 16 36.3% 5 3 2 4 2 Semiconductors & Equipment 6 100.0% 0 1 1 1 3 Software & Services 9 53.3% 6 0 1 0 2 Technology Hardware & Equipment 13 84.5% 6 0 3 1 3 Telecommunication Services 5 4.9% 1 0 0 3 1 Transportation 8 16.1% 4 0 2 2 0 Utilities 6 18.8% 3 0 1 1 1
Source: Smith Barney analysis.
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley86
The BIG Picture
• Theory
•Efficient Frontier
•Asset/Liability Framework
•Alpha/Beta risk budgeting
• Practice
•Pension Crisis
•GM Pension Plan
•Yale Endowment
• Theory
•VaR
•Stochastic behavior of asset
returns
•Worst case scenario
• Practice
•Investment bank VaR
•Bear Stearns
• The Theory
•Active Portfolio Management
•Quantitative equity models
•Hedge fund investment
• The Practice
•Hedge fund industry
•Aug 2007 crisis for quant
Investment BankInstitutional Investor Hedge Fund
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
The backdrop
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Fund flow
88
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Fund flow
89
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Investment landscape before the crisis
90
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Factor volatility and correlation
91
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Investment landscape
Managers both in the long-only but mainly in the HF space, which had been promising double-digit returns to their clients, had to do something. Over the last few years it was quite apparent that three things were happening:
Long-only portfolios became more concentrated and with higher active weights
A large number of 130/30 quant products had been launched
Market neutral hedge funds increased their leverage taking advantage of the very low cost of financing Even the risk disciplined quant investment community had to react to be able to compete with long/short equity managers and survive in a period of very low systematic volatility. But the increase in AUM of Market Neutral Hedge Funds together with multiple levels of leverage applied to their positions gave them significant firing power. adding leverage to their portfolios
92
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Leverage for Market Neutral Hedge funds
93
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Aug 6-Aug 10, 2007
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Aug 6 – Aug 10 2007
95
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Valuation factor daily return
96
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Quality factor daily return
97
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Market sentiment factor daily return
98
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Aug 6-Aug 10, 2007
99
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
A tail event
100
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Market context from Aug 7-Aug 10
101
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Monthly perspective
102
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Hedge fund performance in Aug 2007
103
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Return dispersion
104
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
An experiment
10 out of the 15 largest quantitative managers agreed to give Lehman quant equity research team the rankings coming out of their Core models for 2006 Q1, 2006 Q2 and 2006 Q3 – Top quintile ranked stocks in S&P 500; Bottom quintile ranked stocks in S&P 500 – Top quintile ranked stocks in R2000; Bottom quintile ranked stocks in R2000
We picked a small sample –want homogeneity. These are pure quants. True quants Focus solely on top and bottom quintiles in order to maximize agreement, plus this is where the models take their largest positions
105
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Doing the same things?
106
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Doing the same things?
107
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Quant equity investment landscape
108
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley
Alpha or Beta
109
Financial Risk Management
Prof. Jeff (YuQing) Shen, UC Berkeley110
The BIG Picture
• Theory
•Efficient Frontier
•Asset/Liability Framework
•Alpha/Beta risk budgeting
• Practice
•Pension Crisis
•GM Pension Plan
•Yale Endowment
• Theory
•VaR
•Stochastic behavior of asset
returns
•Worst case scenario
• Practice
•Investment bank VaR
•Bear Stearns
• The Theory
•Active Portfolio Management
•Quantitative equity models
•Hedge fund investment
• The Practice
•Hedge fund industry
•Aug 2007 crisis for quant
Investment BankInstitutional Investor Hedge Fund