Sunteți pe pagina 1din 27

Lecture 10: Discount Factor, Beta, Mean-variance Frontier The following to ics will be covered:

! ! ! ! Discount factor "#$istence Theore% Beta &e resentations Mean-variance Frontier &elation between Discount Factors, Betas, and Mean-variance Frontiers "#'uivalence Theore%( ! )% lications Materials are from chapters 4, 5, & 6, JC.
L10: Discount Factors, beta, and Mean-variance Frontier 1

Law of *ne +rice


+a,off s ace
! The a,off s ace - is the set of all the a,offs that investors can urchase, or it is a subset of the tradable a,offs that is used in a articular stud,. ! +a,off s ace includes so%e set of ri%itive assets, but investors can for% new a,offs b, for%ing ortfolios of the ri%itive assets. This leads to:

+ortfolio for%ation "/1(


! x1, x2 0 X ax1 + bx2 0 X for an, real a, b. ! i.e., 11 to get &, 12 to get 2&, -11 to get -& ! This assu% tion rules out short sell constraint, bid3as4 s read, leverage li%itation, etc. ! 5e can trade nonlinear functions of a basis a,off, e.g., call o tion
L10: Discount Factors, beta, and Mean-variance Frontier 2

Law of *ne +rice


Law of one rice "/2(: p(ax1+bx2) ap(x1) + bp(x2)
! linearit, ! The law of one rice sa,s that investors cannot %a4e instantaneous rofits b, re ac4aging ortfolios. ! The law is %eant to describe a %ar4et that has alread, reached e'uilibriu%. )f there are an, violations of the law of one rice, trader will 'uic4l, eli%inate the% so the, cannot survive in e'uilibriu%

L10: Discount Factors, beta, and Mean-variance Frontier

Theore%: 6iven free ortfolio for%ation /1, and the law of one rice /2, there e$ists a uni'ue a,off x! 0 X such that p(x) "(x!x) for all $.
! x! is a discount factor

Theore% on the #$istence of a Discount Factor

).e., m x! ## a special m p(x) "(mx) "$pro%(m&X)x'

"(x!x)

! x! pro%(m&X) *n the e$istence of % "the discount factor(. 7ni'ue in a,off s ace -. 8ee Figure 9.2 )t can be shown that:

x; = p: " " xx: ( x

L10: Discount Factors, beta, and Mean-variance Frontier

5hat the Theore% Does <ot 8a,

L10: Discount Factors, beta, and Mean-variance Frontier

<o /rbitrage and +ositive Discount Factors <o /rbitrage "absence of arbitrage(
! +ositive a,off i% lies ositive rice. ).e., $=0 then =0.

%=0 i% lies no arbitrage <o arbitrage and the law of one rice i% l, %=0.

L10: Discount Factors, beta, and Mean-variance Frontier

5hat Does )t <ot 8a,>


Discount factor %=0 e$ists, but it does not sa, that %=0 is uni'ue. Does not sa, ever, %=0 5e can e$tend the ricing function defined on - to all ossible a,offs &s

L10: Discount Factors, beta, and Mean-variance Frontier

#$ licit For%ula for $;, the discount factor


/ssu%ing the discount factor $; is the linear function of the shoc4s to a,offs: x! "(x!) + (x ) "(x))*b Finding b to ensure that $; rices the asset $: p 5e have: "(xx!) "(x!)"(x) + "$x(x ) "(x))*'b

b = 1@ p " " x;( " " x(?, (here = " "@ x " " x(?@ x " " x(?:

The convenient alternative for%ula:


x; = " " x;( + @ p " " x;( " " x(?: 1@ x " " x(?
L10: Discount Factors, beta, and Mean-variance Frontier 8

#$ ected &eturn-Beta &e resentation


Linear factor ricing %odel: " " , i ( = + i ,a a + i ,b b + ... i = 1,2,..., + where A ter%s are defined as the coefficients in a %ulti le regression of returns on factors
,ti = ai + i ,a f t a + i ,b f t b + ... + ti t = 1,2,..., -

Factors f are ro$ies for %arginal utilit, growth. The regression is not about redicting return fro% variables seen ahead of ti%e. )ts obBective is to %easure conte% oraneous relations or ris4 e$ osure
L10: Discount Factors, beta, and Mean-variance Frontier 9

#$ ected &eturn-Beta &e resentation


Beta is the a%ount of e$ osure of asset i to factor a ris4s #$ ected return-beta relationshi should be tested via cross-sectional regression
" " , i ( = + i ,a a + i ,b b + ... + i i = 1,2,..., +

Test the constraint:

i = 0

L10: Discount Factors, beta, and Mean-variance Frontier

10

8o%e Co%%on 8 ecial Cases


)f there is a ris4-free rate, we have , f = )f there is no ris4-free rate, then D %ust be esti%ated in the cross-sectional regression. )t is called Eero-beta rate. )n the for% of e$cess returns, we have:
" " , i ( = + i ,a a + i ,b b + ... i = 1,2,..., +

/lso, it is often the case that the factors are returns or e$cess returns.
" " , i ( = + i ,a " " f a ( + i ,b " " f b ( + ... i = 1,2,..., +
11

L10: Discount Factors, beta, and Mean-variance Frontier

Mean-Fariance Frontier
Mean !variance frontier of a given set of assets is the boundar, of the set of %eans and variances of the returns on all ortfolios of the given assets.

L10: Discount Factors, beta, and Mean-variance Frontier

12

Lagrangian /
%ean return: " "(,)

8tart with a vector of asset return &.

roach to 6et Mean-Fariance Frontier

variance-covariance %atri$: . "$(,#")(,#")*' we construct an o tional ortfolio and choose weight ( *bBective: choose a ortfolio to %ini%iEe variance for a given %ean:

%in (: (
(

s.t.

(: " =

(:1 = 1

L10: Discount Factors, beta, and Mean-variance Frontier

13

8olving the 8,ste%


)ntroduce Lagrange %ulti lier on the constrains, we have:
%in (: ( 2 " (: " ( 2 " (:1 1(
(, ,

(: ( = 2( <ote ( is a vector ( 5e have the e$ ression for the variance of the %ini%u% variance ortfolio s ecified in G.H " I2(
! The variance is a 'uadratic function of the %ean.

Mini%u% variance ortfolio has %ini%u% variance, weight s ecified on age IJ

L10: Discount Factors, beta, and Mean-variance Frontier

14

*rthogonal Deco% osition and Mean-Fariance Frontier


/lternative a roach to derive %ean-variance frontier Kansen and &ichard "1LIH( a roach 5e can describe an, return b, a J-wa, orthogonal deco% osition, then the roble% is solved. Define &; -- the return corres onding the a,off $;

x; x; ,; = = p" x;( " " x ;2 (


Define &e;

,; = pro% "1 P , e ( , = space O of O excess O ret/r0s = Nx X



L10: Discount Factors, beta, and Mean-variance Frontier

s.t.

p " x( = 0M
15

*rthogonal Deco% osition and Mean-Fariance Frontier


&e; is an e$cess return that re resents %eans on &e s ace with an inner roduct in the sa%e wa, that $; is a a,off in - s ace that re resents rices with an inner roduct. "(,e) "$pro%(1&,e) ,e' "(,e!,e) Theorem1 "2er3 ,i ca0 be expresse4 as ,i ,! + (i,e! + 0i (here 0i is a0 excess ret/r0 (ith propert3 "(0i) 5 a04 the compo0e0ts are ortho6o0al. Theorem1 ,m2 is o0 the mea0#2aria0ce fro0tier if a04 o0l3 if ,m2 ,! + (i,e! for some real 0/mber (.

L10: Discount Factors, beta, and Mean-variance Frontier

16

Mean-Fariance Frontier in +a,off 8 ace

<ote: &e s ace is the s ace of e$cess returns, thus Q0


L10: Discount Factors, beta, and Mean-variance Frontier 17

*rthogonal Deco% osition in Mean 8tandard Deviation 8 ace

L10: Discount Factors, beta, and Mean-variance Frontier

18

/lgebraic +roof
5ith deco% osition, "(0i) 5 and that the three co% onents are orthogonal, we have:

" " , i ( = " " , ; ( + (i " " , e; (

" , ( = " , + ( , ( + "0 (


2 i 2 ; i e; 2 i

For each desired value of the %ean return, there is a uni'ue (i that %ini%iEe variance for each %ean.

L10: Discount Factors, beta, and Mean-variance Frontier

19

8 anning the Mean-Fariance Frontier


Rou can s an the %ean-variance frontier with an, two ortfolios that on the frontier 7se ris4 free rate or its 4ind to s an the s ace
! Sero-beta return ! Constant-%i%ic4ing ortfolio return ! Mini%u% variance return

8 an and diversification
L10: Discount Factors, beta, and Mean-variance Frontier 20

&;, &e;, $;
8ee age IL through LJ.
1 "1( " " , ; ( = " " x ;2 ( ,; "2( x; = " " , ;2 (
2

"J(

p" x( = " " x ; x( =

" " , ; x( " " , ;2 (

"9( " " , e ( = " " , e; , e (


2 1 " " , ; ( "G( , f = = " " x;( " " , ; (

"T( " " , ; , e; ( = 0 "H( , m2 = , ; + (, e;


L10: Discount Factors, beta, and Mean-variance Frontier 21

&;, & ;, $;
e
"I( &; is the %ini%u% second %o%ent return "L( "(,e!) "(,e!2) "10( )f there is ris4-free rate, ,e! 1 # (17,f)!,! "11( ,f ,! + ,f,e! "12( )f there is no ris4-free rate, 8ro%(1&X) pro%(1&,e)+pro%(1&,!) "1J( x; p : " " xx: ( 1 x

,; =

p " x;(

p : " " xx: ( 1 p

"19(

1 1 p : " " xx : ( x e , ; = 1 f , p: " " xx: ( 1 p


L10: Discount Factors, beta, and Mean-variance Frontier 22

Kansen-Uagannathan Bounds
0 = " "m, e ( = " "m( " " , e ( + m , , e "m( " , e (
e " m ( P " " , (P 5e have, " " m( " , e (

Kansen and Uagannathan "1LL1( read this as a restriction on the set of discount factors that can rice a given set of returns 5e need ver, volatile discount factors with %ean near 1 ""(m) ,f when ris4 free rate e$ists( to understand stoc4 returns. The higher the 8har e ratio, the tighter the bound on the volatilit, of discount factor
L10: Discount Factors, beta, and Mean-variance Frontier 23

#$ licit For%
m = " "m( + @ p " "m( " " x(?: 1@ x " " x(? +

"m( @ p " "m( " " x(?: 1@ x " " x(?

L10: Discount Factors, beta, and Mean-variance Frontier

24

+Q#"%$( A
Beta re resentation using %
cov"m, , i ( var"m( "", ( = + " (" ( = + i ,m m var"m( " " m( (here = 1 3 " "m(
i

Beta re resentation using $; and &; Theore%: 1 "(m,i) i% lies an e$ ected return-beta %odel x! pro%(m&X) or ,! x!7"(x!2) as factors,

" " , i ( = + i , x ; x;
" " , i ( = + i , ,;@ " " ,;( ?
L10: Discount Factors, beta, and Mean-variance Frontier 25

More V
MF Frontier p "(mx) and A " age 10J(
"",i ( = + : i <=> m = a + b: f
"+age 10G(

Discount factors and beta %odels to %ean-variance frontier " age 110(

L10: Discount Factors, beta, and Mean-variance Frontier

26

Factor-Mi%ic4ing +ortfolios
5hen factors are not alread, returns or e$cess returns, it is convenient to e$ ress a beta ricing %odel in ter%s of its factor-%i%ic4ing ortfolios rather than factors the%selves. Defining f! pro%(f&X) m b*f carries the sa%e ricing i% lication on - as does m b*f p "(mx) "(b*fx) "$b*(pro%(f&X)x' "$b*f!x' Factor %i%ic4ing return

pro% " f P X ( f;= p@ pro% " f P X (?

L10: Discount Factors, beta, and Mean-variance Frontier

27

S-ar putea să vă placă și