HEC Paris Financial Markets Fall 2013 Midterm Exam “Cheat Sheet”

Transcription

HEC Paris Financial Markets Fall 2013 Midterm Exam “Cheat Sheet”
HEC Paris
Financial Markets
Fall 2013
Midterm Exam “Cheat Sheet”
0. Basic Statistics
(a) Consider an n-outcome probability space with probabilities p1 , p2 , . . . , pn . Consider two
discrete random variables X and Y with outcomes (X1 , X2 , . . . , Xn ) and (Y1 , Y2 , . . . , Yn ).
The we have the following formulas for means (µX , µY ), variance (σX2 ), standard deviation
(σX ), covariance (σX,Y ), and correlation (ρX,Y )
µX = EX = E(X) = p1 X1 + p2 X2 + · · · + pn Xn
µY = EY = E(Y ) = p1 Y1 + p2 Y2 + · · · + pn Yn
σX2 = var(X) = E (X − µX )2 = p1 (X1 − µX )2 + p2 (X2 − µX )2 + · · · + pn (Xn − µX )2
p
var(X)
σX = σ(X) =
σX,Y = cov(X, Y ) = E (X − µX )(Y − µY )
= p1 (X1 − µX )(Y1 − µY ) + p2 (X2 − µX )(Y2 − µY ) + · · · + pn (Xn − µX )(Yn − µY )
cov(X, Y )
ρX,Y = corr(X, Y ) =
σX σY
(b) Some formulas relating covariances, correlations, standard deviations and variances
cov(X, Y )
cov(a1 X1 + a2 X2 , Y )
var(X1 + X2 )
var(aX + b)
=
=
=
=
corr(X, Y ) σX σY
a1 cov(X1 , Y ) + a2 cov(X2 , Y )
var(X1 ) + var(X2 ) + 2 cov(X1 , X2 )
a2 var(X)
(c) Univariate regression: By regressing the dependent variable Y on the independent (or
explanatory) variable X, one gets the regression line:
Y t = α + β X t + εt ,
where α is the intercept, β is the slope, and εt is the residual (or the error term).
One typically assumes E(εt ) = 0, and cov(Xt , εt ) = 0. The slope β is given by β =
cov(X, Y )/ var(X). The variance of Y decomposes as var(Y ) = β 2 var(X) + var(ε). The
goodness of fit of the regression is measured by R2 = β 2 var(X)/ var(Y ).
1. Present Value
(a) Consider an asset with cash flows Ct+1 , Ct+2 , Ct+3 , . . . If the discount rate r is constant,
the price of the asset is given by the present value formula
Pt =
E (Ct+1 )
E (Ct+2 )
E (Ct+3 )
+
+
+ ...
2
1+r
(1 + r)
(1 + r)3
The discount rate r is the same as the expected return of the asset, and is given by a
model such as CAPM or APT
(b) Similarly, consider a project involving a series of (net) cash flows C0 , C1 , C2 , . . . , CT
occurring in 0, 1, 2, . . . , T periods. The NPV of this project is
NPV = C0 +
C2
CT
C1
+
+ ··· +
2
(1 + r) (1 + r)
(1 + r)T
(c) The future value of a cash flow of C, invested over T periods at a rate of return r is:
FV (C) = C × (1 + r)T
(d) If we know the Annual Percentage Rate (APR), the Effective Annual Rate (EAR), when
interest is compounded each of the m subdivisions of a year, is given by
m
APR
EAR = 1 +
−1
m
(e) If r is the annual rate of return of an investment, it takes T =
72
the investment. The rule of 72 approximates this by T ≈ 100r
ln(2)
ln(1+r)
years to double
(f) The price of a perpetuity that pays C forever (if the discount rate is r) is:
P (Perpetuity) =
C
r
(g) The price of an annuity that pays C for T periods is:
C
1
C
C
P (Annuity) =
−
×
=
T
r
(1 + r)
r
r
1
1−
(1 + r)T
(h) The price of a growing perpetuity that pays initially C and then grows at a rate g per
period forever is:
C
P (Growing Perpetuity) =
r−g
For stocks this is also called the “Gordon dividend growth formula.”
(i) The price of a growing annuity over T periods that pays initially C and then grows at a
rate g is:
C
(1 + g)T
C
C
(1 + g)T
P (Growing Annuity) =
−
×
=
1−
r − g (1 + r)T
r−g
r−g
(1 + r)T
2. Capital Asset Pricing Model (CAPM)
(a) The tangency portfolio T is the market portfolio, with weights given by the market
capitalization of each asset.
(b) The only thing that matters for equilibrium returns is market risk, measured by beta
βi =
cov(˜
ri , r˜M )
2
σM
(c) The relationship between expected returns and beta is the Security Market Line (SML)
Ei = rf + βi (EM − rf )
(d) Beta can be estimated as the regression coefficient in the CAPM regression
e
e
r˜i,t
= αi + βi r˜M,t
+ ε˜i,t
where r˜ie = r˜i − rf is the excess return of security i
(e) The total risk (variance) of asset i can be decomposed
var(˜
r)
| {z i}
=
total risk
βi2 var(˜
r )
| {z M }
+
systematic risk
var(˜
ε)
| {z i}
idiosyncratic risk
(f) The R-squared of the regression is the ratio of systematic risk to the total risk
R2 =
rM )
systematic risk
β 2 var(˜
= i
total risk
var(˜
ri )
3. APT
(a) Assume asset returns depend on k common factors (systematic sources of risk)
r˜i,t = ai +
β1i
|{z}
first factor
loading
× F1t + · · · +
|{z}
first
factor
βki
|{z}
k-th factor
loading
× Fkt +
|{z}
k-th
factor
ε˜i,t
|{z}
idiosyncratic
shock
(b) Then Arbitrage Pricing Theory says that
E i − rf
| {z }
asset i’s
risk premium
=
β1i
|{z}
E T P − rf
| 1{z }
+ ··· +
1st factor 1st factor
loading risk premium
βki
|{z}
ET Pk − rf
| {z }
k-th factor k-th factor
loading risk premium
where T P1 , . . . , T Pk are tracking portfolios: they move one-to-one with the factors
(c) The Fama–French 3-factor model (version of APT)
Ei − rf = βi,MKT EMKT + βi,SMB ESMB + βi,HML EHML
• MKT = Market minus rf ,
• SMB = Small minus Big
• HML = Value minus Growth (High minus Low Book/Market ratio)