Professional Documents
Culture Documents
1.
2.
3.
4.
Session 09
Outline 09: Multifactor Models and Valuation
5. Optimal Portfolios
6. Tailoring Risk Exposures
7. Examples of Factor Models
Value Premium puzzle: firms with high Book-to-MKT ratios (value stocks) perform better
than those with low ratios (growth stocks)
Portfolio Management
Session 09
Portfolio Management
Session 09
R it = ai + bi1F1t + bi 2 F2t + it
t = 1,...,T,
(1)
Portfolio Management
Session 09
(2)
j.
The general form of the corresponding two-factor pricing model, used to represent the sources of risk
premia in expected returns is
Ei = Rf + bi11 + bi22,
(3)
Return Generating
Process
Asset Pricing
Formula
R%it = ai + bi R% Mt + %it
Ei = Rf + i ( Rf)
Ei = Rf + bi11 + bi22
Portfolio Management
Session 09
(4)
If there are no arbitrage opportunities, then there must be a cross-sectional linear relation between Ei and
bi1 of the form
Ei = Rf + 1 bi1,
(5)
Ei
1
Rf
bi1
Copyright Jos M. Marn
Portfolio Management
Session 09
bi1
Example: Suppose that there are three assets with the following expected returns and sensitivities to the
factor:
bi1
Asset (i)
Ei
1
.07
0.5
2
.11
1.5
3
.10
1.0
If we pick any two of these assets and determine the line relating their Eis and bi1s, the other asset does
not lie on that line. For example, the line determined by assets 1 and 2 is
Ei = 0.05 + 0.04 bi1,
Copyright Jos M. Marn
Portfolio Management
Session 09
(6)
(7)
(8)
which is a riskless arbitrage profit the random component of the return, F1t, has been hedged away.
If a riskless asset exists, it too must lie on the line relating Ei to bi1. Since its bi1 is zero, its expected return,
Rf, will be the intercept of the line, as we have written in (5).
More formally:
o Consider the following return generating process for three assets:
Copyright Jos M. Marn
Portfolio Management
Session 09
(a)
(b)
Portfolio Management
Session 09
(c)
o Now, basic knowledge of matrix algebra is enough to conclude that (a), (b) and (c), and x 0 imply
that there exists constants 0 and 1 such that:
Ei = 0 + 1 bi1 i= 1, 2, 3.
o Finally, if one of the assets is the risk free asset, it must be true that 0 = Rf. Hence,
Ei = Rf + 1 bi1
o This completes the proof of the equation (5).
So, for a general factor model:
t = 1,...,T,
(9)
(10)
Portfolio Management
Session 09
Overview of Uses
Applications of multifactor model include
o risk measurement and estimation
o risk management and hedging
o factor-neutral strategies
o trading and arbitrage
o portfolio optimization
o tailoring risk exposures
o style analysis
o performance evaluation
In fixed-income applications, it is often the case that
o the number of factors is small, typically one or two (sometimes three)
o the its are assumed to be zero
o the Fkts (factors) are prespecified as functions of observable variables
o the biks (sensitivities) change over time as functions of the factors
In equity applications, it is often the case that
o the number of factors is greater, typically three or more
o the its are not assumed to be zero
Copyright Jos M. Marn
10
Portfolio Management
Session 09
11
Portfolio Management
Session 09
(11)
(12)
12
Portfolio Management
Session 09
Example:
Suppose an investor instead holds a portfolio or security A, which is correctly priced.
The regression of portfolio As return on the mimicking-portfolio returns gives
~
~
~
R At R f = 0 + 0.5 ( R(1),t R f ) + 0.4 ( R( 2 ),t R f ) + u~At
(13)
13
Portfolio Management
Session 09
By changing the allocations among the K factor-mimicking portfolios, an investors exposure to any of the risk
factors can be increased or reduced.
Examples:
A pension plan sponsor (corporation) whose cash flows are especially sensitive to the business cycle might
wish to have its pension fund take a lower, perhaps negative, exposure to an industrial-production factor. That
way, potential shortfalls in the pension fund would be less likely to occur in periods when the company is less
able to make additional contributions to the fund.
A sponsor of a plan with defined benefits that are indexed to inflation could take more exposure to an inflation
factor.
A financial corporation whose cash flows are especially sensitive to overall fluctuations in credit risk could
take less exposure to such a factor in its pension fund.
In the presence of such hedging concerns, investors do not necessarily attempt to maximize a Sharpe ratio
(INCOMPLETE MARKETS) nor hold portfolios passively.
A multifactor approach allows investors to assess the risk-reward tradeoff for each source of systematic factor
risk.
Implications for equilibrium
The tangent portfolio is some combination of the mimicking portfolios.
Copyright Jos M. Marn
14
Portfolio Management
Session 09
The market portfolio is some, possibly different, combination of the mimicking portfolios.
In other words, the CAPM need not hold.
The relative demands by investors to tilt toward or away from a given source of risk determines E(k) Rf, the
premium for that factor.
If most investors prefer to tilt away from a given factor, then investors willing to tilt toward that factor will
receive a large premium.
15
Portfolio Management
Session 09
ct +1
ct + 2
+
+ ...
1 + k ( 1 + k )2
(14)
One can then think of factors affecting price changes, or asset returns, as factor related either to changes in
expected cash flows or to changes in discount rates.
This reasoning led the authors ultimately to five factors:
1 monthly growth rate of industrial production
2 change in expected inflation
3 unexpected inflation
4 change in the risk premium
5 change in the term structure of interest rates
Copyright Jos M. Marn
16
Portfolio Management
Session 09
The change in the risk premium is measured as the difference in returns between junk bonds (rated below BAA)
and long-term U.S. Government bonds.
The change in the term structure is measured as the difference in returns between long-term U.S. Government
bonds and U.S. Treasury Bills.
The return on the equally weighted portfolio of NYSE stocks is also included as a sixth factor.
17
Portfolio Management
Session 09
the return on small-cap stocks (bottom 50% of firms) minus the return on
large-cap firms (top 50%)
the return on high book-to-market stocks (top 30%) minus the return on low
book-to-market stocks (bottom 30%)
For each asset i, the sensitivities are the slopes (bi, si, hi) in the time-series regression,
Ri,t Rf,t = ai + bi(RM,t Rf,t) + si S M Bt + hi H M Lt + i,t.
(15)
18
for all i
(16)
Portfolio Management
Session 09
When returns or return spreads on portfolios are used instead of non-traded factors, then an equivalent
representation of the pricing relation is
ai = 0
for all i
19
Portfolio Management
Session 09
Descriptor
Formulas
Fundamental
and
Market Data
Risk Index
Formulas
Descriptor
Risk Indices
BARRA
Industry
Allocation
Asset Returns
Monthly
Cross-Sectional
Weighted
Regressions
Estimation
Universe
HICAP
Factor
Loadings
GLS
Weighting
20
Portfolio Management
Session 09
Factor
Returns
Covariance
Matrix
Specific
Returns
Specific
Risk
Forecast
21
Portfolio Management
Session 09
22
Portfolio Management
Session 09
3. Thin Stocks: All stocks that are traded over the counter or are outside the HICAP universe, except those with listed options.
Optioned stocks are distinct for several reasons. First, the option price provides an implicit forecast of the total standard deviation of the
stock itself. Second, optioned stocks tend to be those with greatest investor interest and those with greatest effective trading volume. Stock
trading volume descriptors understate the effective volume because they omit option volume.
The thin stocks, about ten percent of the basic sample, are broken out because they tend to trade differently from other stocks. Over-thecounter stocks and other thinly traded securities show price behavior inconsistent with efficient and timely prices. This stocks also show less
perfect synchrony with market movements, frequent periods in which no meaningful price changes can occur, and occasional outlying price
changes that are promptly reversed. These influences cause some measures of stock prices variability to be biased.
In defining the Variability in Markets index, we standardize the formulas for the three categories relative to one another to provide one index
for the total population.
Variability in Markets Descriptors
A. Optioned Stocks
1. Cumulative Range (+)
2. Beta * Sigma (+)
3. Option Standard Deviation (+)
4. Daily Standard Deviation (+)
B. Listed Stocks
1. Beta *Sigma (+)
2. Cumulative Range (+)
3. Daily Standard Deviation (+)
4. Volume to Variance (+)
5. Log of Price (-)
6. Serial Dependence (+)
7. Annual Share Turnover (-)
Copyright Jos M. Marn
23
Portfolio Management
Session 09
C. Thin Stocks
1. Beta *Sigma (+)
2. Cumulative Range (+)
3. Annual Share Turnover (+)
4. Log of Price (-)
5. Serial Dependence (-)
2. Success (SCS)
The Success index identifies stocks that have been recently successful mainly in terms of stock price but also in terms of earnings. The
relative strength of a stock, shown by the stock price, is the chief indicator of how well the stock has gone. The Success index measures the
success of the company over both the last year and the last five years. It does this in two ways: first, as measured by earnings growth (fiveyear growth in earnings, growth in earnings in the latest year, and present growth in earnings implied by the IBES data); and second, as
measured by price behavior in the market over the last five years and the last year (historical alpha and relative strength). In addition, we use
frequency of dividend cuts as a negative indication.
Success Descriptors
1. Relative Strength (+)
2. Historical Alpha (+)
3. Recent Earnings Change (+)
4. IBES Earnings Growth (+)
5. Dividend Cuts, 5-Year (-)
6. Growth in EPS (+)
24
Portfolio Management