fe 8507 lecture 2 more on dynamic security markets students (1).pptx
TRANSCRIPT
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
1/52
FE 8507 Stochastic Modelling
Asset PricingLecture 2SDF, Arb itrage, Complete Markets and Port
Choice
DR MANDY THAM
NANYANG BUSINESS S CHOOL
AY 2014-2015 MINI-TERM 2
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
2/52
Required
FE8507/DR MANDY THAM/2014_2015
Read Chapter 8, BackRead Lecture 2 notes
Revise Chapter 1 and 4, Back
Read The limits of arbitrage JF article.
Read the Forbes article
Tutorial 2
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
3/52
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
4/52
2.1 More on SDF
Recall ()
.(1)
If a risk-free security is not traded, then in (1) is the shadow risthe zero-beta rate.
(1) is not particularly intuitive until we have replaced the unobservaof some observables.
One way to do so is to choose the explicit form for the utility functio
Lets start with power utility (CRRA)
..(2)
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
5/52
2.1 More on SDF
Recall that + ( ) ..(3)
Now, show that
ln()= (
)
(
)
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
6/52
2.1 More on SDF: Lognormal distribu
A standard statistical result is that if X is lognormal, then
ln ~(, )
(+
)......(5)
where k = con
e.g. ln ~(0.1, 0.25)
2 (.+
.)
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
7/52
2.1 Proofln ~(,
)
(+
)
ln 1
2
Recall +,+ 1
Let X = +,+and conditional on info at time t, also
ln +,+ ln(+,+)
ln(+,+)
(ln(+) (ln(,+) 1
2 ln(+)ln(,+)
(ln(+) (ln(,+)
ln(+
(ln(,+)+ ln +
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
8/52
2.1 ProofNow, let
,+
(money market fund)
ln ln
ln 0
ln + , ln 0
Using (5)
(ln(+) (ln(,+)
ln(+
(ln(,+)+ ln +
We have,
(ln(+)ln() 1
2 ln(+) 0
Rearrange,
ln() (ln(+)
ln(+) .(7)
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
9/52
2.1 ProofNext, take LN of (3)
ln + (+
)
(ln + ) ( +
)
(ln + (
+
)
Finally, sub into (7)
ln() (ln(+) 1
2 ln(+)
(
)
(
) ..(8) Q.E
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
10/52
2.1 Economic IntuitionLet us analyse (8)
ln() ( +
) 1
2(
+
)
measures consumption growth, is interest rate, measures patience and is c
discount factor.
1) Interest rate is high when expected consumption growth is high.High interest rate is required to induce saving today in order to increase consump
2) Interest rate is high whenis low.When people are impatient (is low), a higher interest rate is required to induce
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
11/52
2.1 Economic IntuitionLet us analyse (8)
ln() ( +
) 1
2(
+
)
3) When consumption growth is expected to be volatile, interest rate i You lose more utility as consumption falls than you gain utility from an equal r
consumption. Naturally, you would want to hedge the greater fall in utility by sis a form of precautionary savings.
4) When there is no risk aversion (=0), interest rate depends only on the sufactor (i.e. patience).
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
12/52
2.1 Economic IntuitionLet us analyse (8)
ln() ( +
) 1
2(
+
)
5) As risk aversion () increases, interest rate becomes more responsivconsumption growth. Less willing to deviate from a smooth consumption across time
Eqn (8) indicates correlations and not causality between the LHS and RHS Conversely, we can interpret our correlations in opposite directions.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
13/52
2.1 Economic intuitionLet us analyse (8)
ln() ( +
) 1
2(
+
)
Example:
> 1(more risk-averse than a log-utility with 1), a higher interest rate raiseconsumption more than one-to-one. Substitution effects dominate income effec
Substitution effects => to switch from current consumption to savings.
Income effects=> higher income as real interest rate increases.
0 < < 1(less risk-averse than a log-utility with 1), a higher interest rate rperiod consumption less than one-to-one. Income effects dominate substitution
1(log-utility), a higher interest rate raises second period consumption one-tSubstitution effects equal income effect.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
14/52
2.1 Further remarks
from (8) ln() (
)
(
)
Cross-refer to Week 1- Reading Pennacchi Pg 3
ln() (1 )
(Pennacchi)
Pennacchi define
.
To reconcile our predictions (1)-(5) based on (8) with Pennacchi, let
1 Pennacchi also assumes that +is non-stochastic, therefore
0and
We should use (8) which is the generic case assuming +is stochastic.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
15/52
2.1 Further remarks
FE8507/DR MANDY THAM/2014_2015
In previous analysis, we assume returns and prices are given are in realterms.
+,+ , where,+is the real payoff of asset i at time t
Suppose they are given in nominal terms, we must deflate by the consumer price time t.
+,
,
+
,
, 1
+,+ 1...(9)
Where +=
+
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
16/52
2.2 Hansen-Jagannathan Bounds
FE8507/DR MANDY THAM/2014_2015
Using +,+ 1
+, ,+ +)(,+ 1
Divide by + , +, ,+
+ ,+
1
+
,+ +, ,+ (+)((,+)
,
((,) = +, ,+ ( + ..(10a)Since 1 +, ,+ 1,
| ,
(,)|
= + .(10b) where =
(10b) gives us the Hansen-Jagannathan bounds
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
17/52
2.2 Hansen-Jagannathan Bounds
FE8507/DR MANDY THAM/2014_2015
Sharpe ratio =
,
(,)
Empirical test: The Sharpe ratio of any portfolio/assets should not exceed
Using the same power utility previously, + (
)
((
))
((
))
(
) ( (
) )
(
)
(
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
18/52
2.2 Hansen-Jagannathan Bounds
FE8507/DR MANDY THAM/2014_2015
Assume ln +/ ~(, ),
(
)
+
ln / ~(, (
)
),
(
) +
(
) +
(
) 1
1
1
1 1 .(11) (Hint: use Taylors expans
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
19/52
2.3 CAPMKey assumption of CAPM:
(+) ,,+where k>0
(+), ,+ = k ,+, ,+ = ,+
(+), ,+ = ,+, ,+
Using +,+ 1 (note: This gives the expected return of an asset in eq
+) (,+ (+ , ,+) 1
,+
,,
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
20/52
2.3 CAPM
Since +
()
() , ,+
(),,
()=
,
() (a)
,+ (),,
()=
,,,,
() (b)
(b)/(a) and rearrange,
,+ ,+, ,+
,+[ ,+ ]
,+ [ ,+ ] .(12)
,,,
,and ,+ >0
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
21/52
2.3 CAPMNote:
1
(12) is CAPM and is also known as the Security Market line (SML).
Alpha is actual return minus expected return estimated from an APCAPM.
CAPM is an equilibriummodel because it is derived from +,
which is an equilibrium pricing model.Equilibrium => no mispricing => zero alpha
=> ,
,( )
,
, ,+
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
22/52
2.3 Testing CAPM:
,+
[
,+
]
Test (1): ,+ >0
Test (2): alpha =0
Empirically, CAPM fits the data badly.
Rethink how to specify our SDF ?
Rolls critique
CAPM is untestable.
Joint test:
(1) Empirical market proxy is correct.
(2) CAPM is correct.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
23/52
2.3 CAPM and Hansen-JagannathanBounds
,
(,) = + , ,+ + ..(10a)
Under CAPM, + , ,+ 1
,
(,)=( +
From (10a), we conclude | ,
(,)| +
,
(,)
CAPM specifies an upper limit, namely that any other assets cannotabsolute Sharpe ratio larger than that of the market portfolio.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
24/52
2.4 Arbitrage and the law of one price
An arbitrage opportunity for a finite T horizon is a self-finaprocess such that either
(i)< 0and 0with probability 1 or
(i) 0and 0with probability 1 and > 0witprobability
Cross-reference Chapter 4.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
25/52
2.4 Arbitrage and the law of one price
(i)
< 0and
0with probability 1
A third party sponsored your gamble, and your gamble paid off wsure.
(ii) 0and 0with probability 1 and > 0with
probabilityZero initial investment in the portfolio but the portfolio sometim
positive profit but never with losses.
A special case of arbitrage is when this zero-net investment portfrisklessreturn.
FE8507/DR MANDY THAM/2014_2015
xamp e: r rage
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
26/52
xamp e: r rageOpportunity
Assume no short-sales constraint.
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
27/52
2.4 Arbitrage and the law of one price
If there are no arbitrage opportunities for the finitethen there is a strictly positive SDF process , ,
If there are no arbitrage opportunities for each finitthen there is an infinite horizon strictly positive SDF , , .
Absence of arbitrage => SDF exists => Law of one p
State prices exist for each state => Market is complete.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
28/52
Example: Covered interest parityCovered interest parity condition links spot and forward fo
exchange markets to foreign and domestic money markets uof one price.
:current date 0 forward price for exchanging one unit o
currency T periods in the future. At t=T, you pay dollars of one unit of the foreign currency.
:Spot price of foreign exchange. At t=0, you pay dollaof one unit of the foreign currency.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
29/52
Example: Covered interest parityAt t = 0:
Construct a zero initial investment portfolio by selling forward oneexchange at price . i.e. We are obligated to deliver one unit of fat t=T and receive dollars.
Concurrently, buy
unit of foreign currency and invest
in a foreign
period return .
This will costs us
dollars.
We then borrow
dollars at per-period interest rate .
Net initial investment = 0
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
30/52
Example: Covered interest parityAt t = T:
unit of foreign currency earnings per-period returns over T
unit of foreign currency
We receive dollars for delivery of one unit of foreign currency.
Our debt compounded over T periods:
(
)
Net proceeds: -
()
By law of one price, we must have
.(13
(3) is the covered interest parity condition.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
31/52
2.5 State prices and SDF
If we are willing to assume that there is a finitenumber offuture states of the world, we can express the pricing kerneuseful ways.
Suppose there are kpossible states of the world and nass
Denote the payoff of asset i in state j as.
Let security prices be given by the N-vector P.
Denoting portfolio weights by the N-vector .Cost of the portfolio = P
Payoff of the portfolio =
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
32/52
2.5 State prices and SDFWithin this framework, a state-price vector is defined to be a vectorq=(, )satisfying
(, ) =
.
..(14)
Or simply,
A solution for q exists if X is full-rank(complete market).
=> law of one price holds
=> SDF exists
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
33/52
2.5 State prices and SDFRemember that all uncertaintyin this model is about which state w
the next period.
Let probability of state s = , then from 1 E MR
1 ,= .(15)
Now, we have = ,=
1 ,= (16)
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
34/52
2.5 State prices and SDFCompare (15) and (16),
(15): 1 ,=
(16): 1 ,=
We have
.(17)
( 8)
Thus, the SDF can also be represented by the state-price normalizedprobability of the state occurring.
If exists and is strictly positive, then exist and is strictly positiv
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
35/52
Concept checkUsing (16): 1 ,
=
Derive the present value of an asset that pays out regardless of the realized statein the next period.
What do you call this asset?
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
36/52
Concept checkGiven that states of nature occur randomly. Would you exp
price of a high dividend-yield stock to be higherduring ecrecessionsor during booms?
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
37/52
2.5 State prices and the SDFIn sum,
When state prices exist, the SDF exists.
When the SDF exists, we have the pricing equation 1 Ethe law of one price.
A dynamic security market is said to satisfy the law of on
whenever W and W* are self-financing processequal at any date t.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
38/52
2.6 Limits to Arbitrage
One typically can arbitrage when there are no marNo short-sales constraints
Negligible transaction costs
Depth in the market
However there are limits to arbitrage in the real world.
Read The Limits of Arbitrage
by Andrei Shleifer and Robert W. Vishny, Journal of Finance, Vo(Mar., 1997), pp. 35-55.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
39/52
2.6 Limits to Arbitrage
Read Forbes article: TProof That Tech Is In A
Top chart: Total U.S. scap/GDP.
Bottom chart: S&P50
Suppose you thfair market cap/should be one, warbitrage in yea
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
40/52
2.7 Risk-Neutral Probability/PricingWe can perform the same trick that produced a pricing ke
sort of probability" that we care to consider. Many results sometimes called risk-neutralprobabilities.
Using ,=
, ,,
=
, ,
= (20)
where ,
has the characteristics of probability.
1. 0
2. = ,
= ,
, 1
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
41/52
2.7 Risk-Neutral Probability/PricingWe can then write
,
,= =
,
(,)
Since we have
,
Under risk-neutral pricing, we again have
, , =
, (21)
We have changed the probability measure from the statistprobabilityto the risk-neutral probability .
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
42/52
2.7 Risk-Neutral Probability/PricingFormally,
Suppose a money market account exists. Index this as ass
free return from date t to t+1 is,
,
A risk-neutral probability for < is defined to be a promeasure Q where
1. For any event A, (A)=0 if and only if Q(A)=0.2. For any asset i=2,n,
is a martingale on the time horizon
relative to Q.
If 1 and 2 hold, and Q are called Equivalent Martingale Me
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
43/52
2.7 Risk-Neutral Probability/PricingThus,
,
,
,
,
(,)
,
,
,= , (22)
Under risk-neutral probability Q, the expected return on each asset freereturn.
Sub
,
,
,+,
, into (22)
We have
(,+,)
, ,(23)
To compute the price at t, compute the expectation of its value at t+by risk-free return.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
44/52
2.7 Risk-Neutral Probability/Pricing
The mapping from -measure to Q-measure is called the
Nikodymderivative of Q with respect to , denoted by
Think of Radon-Nikodymderivative as the ratio of probabunder Q relative to .Now you should understand why Q and must be equivalent mea
Valuation via a risk-neutral probability is equivalent toan SDF process.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
45/52
2.7 Risk-Neutral Probability/PricingWe have
,
, ..(24)
Proof:
Recall , and ,= ,
Therefore,
,
, =
,= ,,
=
,
= =
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
46/52
2.7 Risk-Neutral Probability/PricingWe have
,
,1 ..(25)
Proof:
Recall ,Since ,= , =
,, ( ,,
= 1) ,,1
where 1is an indicator function that takes on unity only when s=
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
47/52
2.7 Risk-Neutral Probability/PricingWe also have
,
,
..(26)
Proof:
Recall ,
Since ,=,
,
,
,
,
,,
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
48/52
2.8 Complete Markets1) A dynamic securities market is said to be completeif, for eve
random variable x that depends only on date-t information isdate-t payoff.
A random variable x is marketed if it is the payoff of some port
Recall
(, ) =
.
(1) means that []is linearly spanned by a set of marketed portfoliwords, [] is full rank and can be invested to find q.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
49/52
2.8 Complete Markets
If the market is completeand law of one price holds, thencan be at most one SDF processbecause the condition
implies that =, since both
depend on date-t information.
If the market is completeand there is no arbitrageopportunities, then there is a uniqueSDF process, and it is positive.
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
50/52
2.9 Portfolio Choice in Complete MarkeSuppose an investor can choose any consumption process C satisfying
[ ]= ..(27)
Intertemporal budget constraint with no endowment is given by
+ ( )+
Using CRRA utility
()
Recall (
(
=>+
..(28)
Iterating on (28), we have
..(29)
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
51/52
2.9 Portfolio Choice in Complete MarkeFrom (27): [ ]
=
Sub (29), we have
[
]= [
]=
Which can be solved as
+[
]
.(30)
Sub (30) into (29),
+[
]
(3
FE8507/DR MANDY THAM/2014_2015
-
8/11/2019 FE 8507 Lecture 2 More on dynamic security markets students (1).pptx
52/52
Next week
Tutorial 2 presentation
Dynamic Programming, Chap 9, Back.
FE8507/DR MANDY THAM/2014_2015