luis a. seco sigma analysis & management … risk...luis a. seco sigma analysis & management...
TRANSCRIPT
Slide 1
Investment risk management Traditional and alternative products
Luis A. Seco Sigma Analysis & Management
University of Toronto RiskLab
Slide 2
A hedge fund example
Slide 3
A hedge fund example
Slide 4
A hedge fund example
Slide 5
A hedge fund example
Slide 6
A hedge fund example
Slide 7
The snow swap
! Track the snow precipitation in late fall and early spring;
! If the precipitation is high, the ski resort pays to the City of Montreal a prescribed amount.
! If the precipitation is low, the City pays the resort another pre-determined amount.
! The dealer keeps a percentage of the cash flows.
Slide 8
A hedge fund example
The snow swap
© Luis Seco. Not to be distributed without permission.
City Resort
Snow
No snow
$10M
The snow fund
! Modify the snow swap so the City pays when precipitation is low in the city, and the resort pays when precipitation is high in the resort.
! The fund takes the “spread risk”, and earns a fee for the risk. ! Say the “insurance claim” is $1M. The fund would charge 20%
commission, but assume to take the spread risk. ! Setting aside $2M, and charging $200K, the fund could
– Lose nothing: 75% – Make $2M: 12.5% – Lose $2M: 12.5%
! Expected return=10%. Std=50%
A diversified fund: a hedge fund.
! If we do the swap across 100 Canadian cities:
! Expected return:10% ! Std: 5%. ! Better than investing in
the stock market.
So we create do the snow investment…
… with some of the best known ski resorts:
! Blue Mountain (Toronto) ! Mountain Creek (New Jersey) ! Panorama Mountain Village (Calgary) ! Snowshoe Mountain (West Virginia) ! Steamboat Ski Resort (Hayden, Denver) ! Stratton Mountain Resort (Vermont) ! Tremblant (Montreal) ! Whistler Blackcomb (Vancouver)
… and then:
© Luis Seco. Not to be distributed without permission.
Intrawest goes Bankrupt
© Luis Seco. Not to be distributed without permission.
Slide 14
Hedge Fund: definition
! An investment partnership; seeks return niches by taking risks, which they may hedge or diversify away (or not).
! Unregulated ! Bound to an Offering Memorandum ! Seeks returns independent of market
movements ! Reports NAV monthly ! Charges Fees: 1-20
Slide 15
The investment structure
The Management company “the hedge fund”
The Fund legal structure
The Bank Prime Broker The Administrator
Investor 1 Investor 2 Investor 3 Investor 4 Investor n
Slide 16
Risks per strategy
© Luis Seco. Not for dissemination without permission.
Slide 17
Slide 18
Convertible arbitrage
Fig. 1: A graphical analysis of a convertible bond. The different colors indicate different exercise strategies of call and put options.
Risk management for financial institutions (S. Jaschke, O. Reiß, J. Schoenmakers, V. Spokoiny, J.-H. Zacharias-Langhans).
The Galmer Arbitrage GT
Slide 19
Convertible arbitrage
! The convertible arbitrage strategy uses convertible bonds.
! Hedge: shorting the underlying common stock. ! Quantitative valuations are overlaid with credit and
fundamental analysis to further reduce risk and increase potential returns.
! Growth companies with volatile stocks, paying little or no dividend, with stable to improving credits and below investment grade bond ratings.
Slide 20
An convertible arbitrage strategy example
! Consider a bond selling below par, at $80.00. It has a coupon of $4.00, a maturity date in ten years, and a conversion feature of 10 common shares prior to maturity. The current market price per share is $7.00.
! The client supplies the $80.00 to the investment manager, who purchases the bond, and immediately borrows ten common shares from a financial institution (at a yearly cost of 1% of the current market value of the shares), sells these shares for $70.00, and invests the $70.00 in T-bills, which yield 4% per year. The cost of selling these common shares and buying them back again after one year is also 1% of the current market value.
Slide 21
Scenario 1
Values of shares and bonds are unchanged:
Today 1 yr later Bonds 80 80 Stock -70 -70 T-Bill +70 +72.8 Coupon 4 Fee -3.5 Total $80 $83.3
Slide 22
Scenario set 2
In the next two examples, the share price has dropped to $6.00, and the bond price has dropped to either $73.00 or $70.00, depending on the reason for the drop in share market values. The net gain to the client is 7.87% and 4.12% respectively, again after deducting costs and fees.
Today 1 yr later (a) 1 yr later (b)
Bonds 80 73 70 Stock -70 -60 -60 T-Bill +70 +72.8 72.8 Coupon 4 4 Fee -3.5 -3.5 Total $80 $86.3 $83.3
Slide 23
Scenario set 3
In the following three examples, the share price increased to $8.00, and the bond price increased either to $91.00, $88.00 or $85.00, depending on the expectations of investors, keeping in mind that we have one less year to maturity. The net gain to the client is 5.37% and 1% in the first two examples, with an unlikely net loss of 2.12% in the last example.
Today 1 yr later(a) 1 yr later(b) 1 yr later(c)
Bonds 80 91 88 85 Stock -70 -80 -80 -80 T-Bill +70 +72.8 +72.8 +72.8 Coupon 4 4 4 Fee -3.5 -3.5 -3.5 Total $80 $84.3 $81.3 $78.3
Slide 24
A Risk Calculation: normal returns
If returns are normal, assume the following:
Bond mean return: 10% Equity mean return: 5% Libor: 4% Bond/equity covariance matrix
(50% correlation):
! Mean return (gross): 10-5+4=9%
! Standard deviation:
Slide 25
Long-short equity
William Holbrook Beard (1824-1900)
Slide 26
A long-short pair trade
! The fund has $1000. The manager is going to purchase stock 9 units of stock A, and sell-short 9 units of stock B. Both are valued at $100 each. After a year, A is worth $110, B is $105.
Assets at Prime Broker
(Before trade)
• $1000
Assets at Prime Broker
(After trade)
• $1000
• -$900 + 9 A
• +$900 – 9 B
Assets at Prime Broker
(After one year)
• $1000
• 990
• -945
• -9
$ 1036
Slide 27
A long-short pair trade (v2)
! The fund has $500. The manager is going to purchase stock 9 units of stock A, and sell-short 9 units of stock B. Both are valued at $100 each. After a year, A is worth $110, B is $105.
Assets at Prime Broker
(Before trade)
• $500
Assets at Prime Broker
(After trade)
• $500
• -$900 + 9 A
• +$900 – 9 B
Assets at Prime Broker
(After one year)
• $500
• 990
• -945
• -9
$ 536
Slide 28
A long-short pair trade (v3)
! Assumptions: 50% collateral for long trades, 80% collateral for short trades.
Securities at Prime Broker
• 9 A ($900):
• – 9 B (-$900):
Collateral required:
$450+$720=$1170
Cash from short sale: $900
Cash required: $270
Securities at Prime Broker
• 9 A ($990):
• – 9 B (-$945):
Profit: $36
Slide 29
Risk and Performance Measurement
Slide 30
Measurement
! Return: – from track records
! Risks: – Volatility – Operational risk: due diligence – Business risk – Exposures to market factors
© Luis Seco. Not to be reproduced without permission
Slide 31
Return
! Starting from share value observations Si on a monthly basis, we define the return as
! Simple Returns: Ri = (Si - Si-1)/Si-1
! Log Returns Ri = ln(Si/Si-1)
© Luis Seco. Not to be reproduced without permission
Slide 32
TWR and IRR
! Over a period of time, the time-weighted-rate of return is defined by
1+TWR = (1+R1)(1+R2)… (1+Rk) ! Over the same period of time, the Internal Rate of
Return is defined as IRR=(1+R)n
where the number R is defined as
and Ni denote the cashflows at month i.
Return statistics
© Luis Seco. Not to be reproduced without permission
Slide 33
Heaven (Probability)
! Assumes a probability distribution
! Assumes total knowledge
! Expressed with mathematical formulas
Earth (Statistics)
! Derives distributions from history
! Only knows the past
! Implementable on a computer
Slide 34
The portfolio distribution function (CDF)
90% probability that annual returns are less than 3%
7% probability that annual losses exceed 5%
Slide 35
Probability density: histogram
Slide 36
Mean Return
! Return is usually measured on a monthly basis, and quoted on an annualized basis.
! If the series of monthly returns (in percentages) is given by numbers ri, where the subindex i denotes every consecutive month, the average monthly return is given by
! Because returns are expressed in percentages, one has to be careful, as the following example shows.
Mean return estimators
© Luis Seco. Not to be reproduced without permission
Slide 37
! Heaven (Probability) ! Earth (Statistics)
Usually measured monthly, and reported annually
Slide 38
Returns: careful.
Imagine a hedge fund with a monthly NAV given by
$1, $2, $1, $2, $1, $2, etc. The monthly return series is given by 100%, -50%, 100%, -50%, 100%, -50%, etc. Its average return (say, after one year) is 25%
monthly, or an annualized return in excess of 300%.
Slide 39
Returns: from monthly to annual
There is no standard method of quoting annualized returns:
One possibility is multiplying returns by 12 (annual return with monthly compounding)
Another, is to annualize using the formula
Slide 40
Portfolio returns
The big advantage of “return”, is that the return of a portfolio is the average of the returns of its constituents.
More precisely, if a portfolio has investments with returns given by
with percentage allocations given by
then, the return of the portfolio is given by
Slide 41
Volatility
! Like returns, volatility is usually measured on a monthly basis, and quoted on an annual basis.
! If the series of monthly returns (in percentages) is given by numbers ri, where the subindex i denotes every consecutive month, the monthly volatility is given by
Volatility
© Luis Seco. Not to be reproduced without permission
Slide 42
! Heaven (Probability) ! Earth (Statistics)
€
ˆ µ = r = 1n
rii=1
n
∑
σ =1n
(ri − µ)2
i=1
n
∑
s =1
n −1(ri − r )2
i=1
n
∑ Sample s.d.
Population s.d.
Slide 43
Covariances and correlations
! They measure the joint dependence of uncertain returns. They are applied to pairs of investments.
! If two investments have monthly return series given by numbers ri and si respectively, where the subindex i denotes every consecutive month, and their average returns are given by r and s, their covariance is given by
! If they have volatilities given respectively by
! Then, their correlation is given by
Slide 44
Covariance and correlation matrices
Because correlations and covariances are expressed in terms of pairs of investments, they are usually arranged in matrix form.
If we are given a collection of investments, indexed by i, then the matrix will have the form
© Luis Seco. Not to be reproduced without permission
Slide 45
Fund-of-Fund Risk: volatility
Volatility of a portfolio with weights w
Slide 46
Portfolio Optimization: Markowitz
Markowitz optimization allows investors to construct portfolios with optimal risk/return characteristics.
Risk is represented by the portfolio expected return
Risk is represented by the standard deviation of returns.
The optimization problem thus created is LQ, it is solved using standard techniques.
Slide 47
Risk/return space
A portfolio is represented by a vector θ which represents the number of units it holds in a vector of securities given by S.
Each security Si is assumed a gaussian return profile, with mean µi, and standard deviation given by σi. Correlations are given by a variance/covariance matrix V.
The portfolio return is represented by its return mean
and its risk is given by its standard deviation
Slide 48
The efficient frontier
Risk
Return
Feasible
Region
Efficient Portfolios
Slide 49
Sharpe’s ratio
A way to bring return and risk into one number is by the information ratio, and by the Sharpe’s ratio.
If a certain investment has a return given by r, and a volatility given by σ, then the information ratio is given by r/ σ.
If interest rates are given by i, then Sharpe’s ratio is given by (r-i)/ σ.
It measures the average excess return per unit of risk. Portfolios with higher Sharpe’s ratios are usually better.
Slide 50
Sharpe’s ratio: basic fact
! Imagine one is looking for the portfolio that has the best chance of optimizing its performance against a benchmark given by LIBOR. That portfolio is the one with the highest Sharpe ratio, as defined in the previous paragraph.
Slide 51
Sharpe Ratio
The objective function to maximize is
Since φ is increasing, our optimization problem becomes that of maximizing
Probability of meeting the benchmark
Cummulative distribution function
of the gaussian
Slide 52
Sharpe vs. Markowitz
Slide 53
Benchmarks
They are reference portfolios against which performance of other portfolios are measured:
! Bonuses are paid on benchmark-based performance.
! They can be constant or random
Slide 54
Tracking error
! It is the standard deviation of the difference between the portfolio returns and the benchmark returns.
! A performance indicator often times used in traditional investments is
Slide 55
Alpha and beta
Consider a portfolio with returns given by
and a benchmark with returns given by.
Find the linear regression coefficients α, β, such that,
with ε with mean 0 and lowest standard deviation.
Slide 56
VaR and risk budgeting
Assume a portfolio represented by a vector θ which represents the percentage allocated to specific managers or investment instruments.
Each manager or security Si is assumed a gaussian return profile, with mean µi, and standard deviation given by σi. Correlations are given by a variance/covariance matrix V.
VaR and portfolio standard deviation are related to the fundamental expression
Slide 57
Risk budgeting
The previous expression allows us to do a risk allocation to each manager
in such a way that the overall risk of the portfolio is given by
This expression is useful when allocating risk or risk limits to each of the investments in a certain universe.
Slide 58
The normality assumption
Under the normal assumption, a portfolio with a 1% standard deviation will have annual returns which will vary no more than 1%, up or down, from its expected return, with a 65% probability.
If a higher degree of certainty about portfolio performance is desired, then one can say that the portfolio return will vary more than 2% from its expected return only 1% of the times.
These probabilities are linear in the standard deviation; in other words, if the portfolio volatility is 3% (instead of 1% as in the example above), one will expect the returns to oscillate within a 6% band of its average return 99% of the time.
© Luis Seco. Not to be reproduced without permission
Slide 59
Non-normal returns
Slide 60
Gain/loss deviation
It measures the deviation of portfolio returns from its expected return, taking into account only gains. In other words, portfolio losses are not taken into account with calculating the deviation.
Loss deviation is the corresponding thing when losses only are taken into account in calculating portfolio deviations.
Both of these are used when one is trying to get a feeling as to the asymmetry of the gain/loss distribution. They are not statistically conclusive amounts per se, like standard deviation is.
Slide 61
Semi-standard deviation formula
Target return / benchmark
Gains give a value ot 0
Slide 62
Sortino ratio
It is the substitute of the Sharpe ratio when one looks only at the loss deviation, instead of looking at the combined standard deviation.
Many people believe that by not punishing unusual gains, like the Sharpe ratio does indirectly, one maximizes the upside while maintaining the downside.
There is however no evidence that the Sortino ratio, as such, actually achieves this but it still remains to be a curious quantity to look at.
Slide 63
Moments
One of the criticisms of the use of volatilities and correlations as risk measures is the presence of extreme events in portfolio returns, which will go un-noticed in those calculations.
From a certain viewpoint, volatilities and correlations are magnitudes inherited from normal distributions, according to which events such as the ones in 1987, 1995, 1998, etc. should have never occurred.
One attempt to capture “tail events” is by introducing higher moments to measure large deviations: higher moments are defined as follows:
Slide 64
Skew and kurtosis
! Skew is a measure of asymmetry. It is the normalized third moment.
! Kurtosis is a measure of spread. It is the fourth moment, minus 3. Platykurtotic: k<0 Leptokurtotic: k>0 Mesokurtotic: k=0.
Slide 65
Slide 66
Slide 67
Biased estimators
! The estimator for the skewness and kurtosis introduced earlier is biased: – Its expected value can even have the opposite sign from the true
skewness (or kurtosis).
! Intuitively speaking, the third and fourth powers are so large, that one or two events will dominate the value of the formula, making all other observations irrelevant.
! Skew and kurtosis should not be used in critical situations
Slide 68
Skewness is useless
Slide 69
Uselessness of skewness
Slide 70
L-moments
Slide 71
The Omega
Slide 72
Omega
! Shadwick introduced the concept of “Omega” a few years ago, as the replacement of the Sharpe ratio when returns are not normally distributed.
! His aim was to capture the “fat tail” behavior of fund returns.
! Once the “fat tail” behavior has been captured, one then needs to optimize investment portfolios to maximize the upside, while controlling the downside.
Omega: Shadwick, Keating (2002)
Slide 73
Slide 74
Wins vs. losses: the Omega
Omega tries to capture tail behavior avoiding moments, using the relative proportion of wins over losses:
Slide 75
Wins vs. losses: the Omega
Omega tries to capture tail behavior avoiding moments, using the relative proportion of wins over losses:
Truncated First Moments
Slide 76
The Omega of a heavy tailed distribution
Correlation risk
Slide 77
Slide 78
Hedge fund diversification
Hedge funds are uncorrelated to traditional markets, and internally uncorrelated also.
Correlation histogram for Dow stocks
Correlation histogram for hedge funds
Slide 79
Fact.
Hedge funds are uncorrelated to traditional markets, so they constitute excellent diversification strategies.
Yes, ... and no! Many hedge funds are indeed
uncorrelated to markets, but others are very correlated to simple portfolios of traditional markets, so they add little diversification.
Even those funds which exhibit low correlation to markets and macroeconomic factors, when combined into portfolios, they can be highly correlated to the market.
Slide 80
Hedge Fund Correlation histogram
Slide 81
Normal correlations
Slide 82
Distressed correlations
Slide 83
Correlation switching
Slide 84
Distress analysis
Slide 85
Correlation switching
Slide 86
Correlation risk
We will deal with correlation sensitivity from a mixtures of multivariate gaussian approach
Its density is given by:
Slide 87
GM in pictures
Slide 88
Non-gaussian portfolio theory
Each portfolio is described by four performance numbers: mean and standard deviation, each under normal and distressed market assumptions. They are given by
and
Slide 89
Benchmark satisfaction
The objective function to maximize was
It is possible to have portfolios which are efficient from this point of view, which however are not efficient under either normal or distressed conditions.
Increasing functions
Slide 90
Other risks
! Backfill bias ! Survivorship bias ! Liquidity risk ! Style risk ! Legal risk ! Non-linear effects: option writing.
Slide 91
Hedge Fund Products
! Fund-of-funds: Indices ! Options on fund-of-funds ! Warrants ! Non-recourse loans with fund collaterals ! CPPI (Constant proportion portfolio
insurance) ! CFO’s
Slide 92
Hedge Fund indices
! They offer fund-of-fund investments that try to track the performance of the hedge fund sector (global and style specific) investing in liquid funds with high capacity.
! The result is a fund that tracks nothing and lags performance.
! In contrast with equity indices, investors in a fund don’t like it when their fund is included in an index.
Slide 93
Hedge Fund Indices
! Investable ! Non-investable
Slide 94
Historical comparative analysis
Pro-Forma
Slide 95
Correlation analysis
Slide 96
Guaranteed notes
! There are two main reasons for a guarantee: – Regulatory environments – Risk perceptions (not to confuse with risk appetite)
! Some guarantees are provided by well-rated banks. Others are not (Portus).
! Guarantees are obtainable by setting aside an interest-earning portion of the assets, and investing the remainder at higher levels of leverage, through a variety of different instruments.
Slide 97
Anatomy of a guarantee Guarantees principal in
the future: How much is needed is determined by
• Interest rates
• Maturity date of the note
Obtains exposure to the Hedge Funds
Slide 98
The cost of the guarantee
About 2% per year cost
An underlying hedge fund portfolio that produces 6bps/month
Interest rates at 25bps per month A 5 year note that guarantees principal No management or performance fees
Leveraged structures
Loans Options
CPPI
Slide 99
Slide 100
Non-recourse loans
! The bank lends to the investor and takes the investment in the hedge fund portfolio as collateral.
! In a low interest rate environment, it allows investors to amplify good hedge fund performance. In high interest rate environments, if hedge fund performance is poor, they can lead to sustained losses.
! It allows small investors to increase the asset base and diversify the portfolio better; it makes it easier to satisfy the minimum investment requirements of individual hedge funds.
! The structurer may demand liquidation if performance drops below a certain floor.
Slide 101
Options
! Options are delta-hedged; the liquidity of the underlying hedge fund portfolio contributes to a volatility spread.
! They are hard to delta-hedge due to the low liquidity of the underlying portfolio. Implied volatilities will be much higher than historical volatilities.
! They are path-independent. They are also insensitive to changes in interest rates.
Slide 102
CPPI
! Investor provide equity to a fund; ! the structurer provides leverage ! Proceeds are invested in a reference portfolio ! If the performance of the reference portfolio is
below a reference curve, the strike price is increased.
! If performance of the reference portfolio is above another reference curve, the strike price is decreased
Slide 103
CPPI options
Slide 104
Bank
Bond Investor (1)
Bond Investor (2)
Bond Investor (3)
Equity Investor
Fund Pool
Collateralized Fund Obligation (CFO)
Slide 105
A $500M CFO
Slide 106
CFO’s
Advantages ! Equity investors find a way
to obtain leverage. ! Debt holders find an
uncorrelated asset class to invest in.
! Tranches can be packaged by volume and credit rating.
Disadvantages ! Hard to value ! Very dependent on
correlations amongst the funds constituents
! Expensive structuring fees makes it difficult to find the equity investor sometimes.
Slide 107
S&P CTA CFO. A case study.
© Luis Seco. Not to be reproduced without permission
Slide 108
Blow-up risk
© Luis Seco. Not to be reproduced without permission
Slide 109
The Merton model of default
Slide 110
A double-layer rating system
A B C
A Infrequent, small losses
Frequent, small losses
B
C Infrequent, large losses
Large, probably losses
Slide 111
Rating and Due Diligence