This action might not be possible to undo. Are you sure you want to continue?

Eckhard Platen1

and

Wolfgang J. Runggaldier2

March 11, 2002

Abstract. The paper proposes the use of the growth optimal

portfolio for the construction of nancial market models with unobserved factors that have to be ltered. This benchmark approach avoids any measure transformation for the pricing of derivatives. The suggested framework allows to measure the reduction of the variance of derivative prices for increasing degrees of available information.

1991 Mathematics Subject Classi cation: primary 90A09 secondary 60G99, 62P20. JEL Classi cation: G10, G13 Key words and phrase: Financial modelling, lter methods, benchmark approach, growth optimal portfolio.

1

University of Technology Sydney, School of Finance & Economics and Department of Mathematical Sciences, PO Box 123, Broadway, NSW, 2007, Australia 2 Universita degli Studi di Padova, Dipartimento di Matematica, Pura ed Applicata Via Belzoni, 7 I - 35131 Padova, Italy

1 Introduction

In nancial modelling it is sometimes the case that not all quantities, which determine the dynamics of security prices, can be fully observed. Some of the factors that characterize the evolution of the market are hidden. However, these unobserved factors may be essential to re ect in a market model the type of dynamics that one empirically observes. This leads naturally to lter methods. These methods determine the distribution, called lter distribution, of the unobserved factors, given the available information. This distribution allows then to compute the expectation of quantities that are dependent on unobserved factors, for instance, derivative prices. There is a growing literature in the area of ltering in nance. To mention a few recent publications let us list Elliott & van der Hoek (1997), Fischer, Platen & Runggaldier (1999), Elliott, Fischer & Platen (1999), Fischer & Platen (1999), Landen (2000), Gombani & Runggaldier (2001), Frey & Runggaldier (2001), Elliott & Platen (2001), Bhar, Chiarella & Runggaldier (2001a, 2001b) and Chiarella, Pasquali & Runggaldier (2001). All these papers provide examples, where lter methods have been applied in the area of nance. Such applications involve optimal asset allocation, interest rate term structure modelling, estimation of risk premia, volatility estimation and hedging under partial observation. A key problem that arises in most ltering applications in nance is the determination of a suitable risk neutral equivalent martingale measure for the pricing of derivatives. The resulting derivative prices and hedging strategies depend often signi cantly on the chosen measure. On the other hand it is obvious that in ltering one has to deal with the real world probability measure. It is therefore important to explore alternative methods that are based on the real world measure and allow consistent derivative pricing. In this paper we suggest a benchmark approach to ltering, where the benchmark portfolio is chosen as the growth optimal portfolio (GOP), see Long (1990) and Platen (2002). The GOP has the important economic interpretation of being the portfolio that maximizes expected logarithmic utility. The dynamics of the growth optimal portfolio depends on the degree of available information. Given a certain information structure, one naturally obtains in this approach a fair price system, where benchmarked prices equal their expected future benchmarked prices. This avoids the involvement of a risk neutral equivalent martingale measure. All resulting prices, when expressed in units of the GOP, turn out to be local martingales under the given real world measure. In cases when benchmarked prices are strict local martingales the benchmark approach generalizes the standard risk neutral approach. The paper is structured in the following way. It summarizes in Section 2 the general ltering methodology for multi-factor jump di usion models with unobserved factors. Section 3 describes the proposed ltered benchmark model. The 2

3) for t 2 0 T ] with given initial value z0 . 2 Filtered Multi-Factor Models 2. Here w = wt = wt : : : 1 n . while the remaining n .2) with yt representing the observed and xt the unobserved factors.(yt.fair pricing and hedging of derivatives is then studied in Section 4.k )> = (ztk+1 : : : ztn)> t t (2. Here k is an integer with 1 k < n that we shall suppose to be xed during most of this paper. yt may represent the vector of logarithms of the continuous and jump parts of observed risky security prices. wtk . only the rst k factors are directly observed.) dmt dyt = At (zt ) dt + Bt(yt) dvt + Gt. We consider a multi-factor model with n 2 factors z1 z2 : : : zn . forming the vector process z = zt = zt1 : : : ztk ztk+1 : : : ztn n . wtk+1 ::: wtn > > t 2 0 T] o (2. (zt.1 Factor Model To build a nancial market model with a su ciently rich structure and high computational tractability we introduce a multi-factor model.1) We shall assume that not all of the factors are observed. We assume that the observed and unobserved factors satisfy the system of stochastic di erential equations (SDEs) dxt = at (zt ) dt + bt (zt ) dwt + gt.4 we shall discuss the implications of a varying k. This section also demonstrates how to quantify the reduction of the variance for derivative prices using more information.5) is an n-dimensional (A P )-Wiener process and vt = wt1 : : : wtk 3 . k are not. For xed k we shall consider the following subvectors of zt yt = (yt1 : : : ytk )> = (zt1 : : : ztk )> and xt = (x1 : : : xn. > t 2 0 T] : o (2.) dNt (2. For instance. in Section 4. where A = (At)t2 0 T ] is a given ltration to which all the processes will be adapted.4) (2. However. Let there be given a ltered probability space ( AT A P ). This model provides the basis for the dynamics of nancial quantities. More precisely.

since there are no common jumps among the components of Nt. De ne the ith jump martingale by dmit = dNti . The process m = fmt = (m1 : : : mk . we assume that the vectors at (zt ). see Protter (1990). invertible for each t 2 0 T ]. In addition to the ltration A.6) t t t t This means.9) 1 k ~t where Ak = fys = (zs : : : zs )> s tg represents the observed information k provides the structure of the actually available ~ at time t 2 0 T ]. ~t We shall be interested in the conditional distribution of xt . by observing a jump of yt we can establish which of the processes N i i 2 f1 2 : : : kg.8) Nt = Nt1 : : : Ntk . Let (2. we shall also consider the sub ltration ~ ~t Ak = (Ak )t2 0 T ] A (2. that. which represents the complete information.3) exists that does not explode until time T . Thus A information in the market. We shall also assume that the k k-matrix Bt (yt) is invertible for all t 2 0 T ]. where for i 2 f1 2 : : : kg t t i (z ) = ~ i (y ): (2. t 2 0 T ]g is an n-dimensional (A P )-jump martingale de ned t t as follows: Consider n counting processes N 1 : : : N n having no common jumps. a bounded exp { xt ] is. in particular. given Ak . it (zt ) dt for t 2 0 T ] and i 2 f1 2 : : : ng. Notice that. we assume without loss of generality that the jump intensities of the rst k counting processes are observed. gt(zt ) and Gt(yt) are such that a unique strong solution of (2. Bt(yt).3).3). There exist general lter equations for the dynamics described by the SDEs in (2. k and with { denoting the imaginary unit. has jumped. which depends on the speci cation of the degree of available information k. > be the vector of the rst k counting processes at time t 2 0 T ]. This latter assumption implies that. given Ak . given Ak . It turns out that these are SDEs for the conditional expectations of inte~t grable functions of the unobserved factors xt . Finally. At (zt).7) (2. for given 2 < and thus integrable function of xt . t (zt ) and the matrices bt (zt ). according to standard terminology we call the lter distribution at time t 2 0 T ].is the subvector of its rst k components. t t mk+1 : : : mn)>. These are at time t 2 0 T ] characterized by the corresponding vector of intensities t (zt ) = ( 1(zt ) : : : n(zt ))>. Its conditional expectation leads therefore to ~t the conditional characteristic function of the distribution of xt . Concerning the coe cients in the SDE (2. gt(zt ) may be any bounded function and the k k-matrix Gt(yt ) is assumed to be a given function of yt. The 4 .

In the following two subsections we recall two classical such models. ~t P ztk+1 zk+1 : : : ztn zn Ak = Fz +1 ::: z zk+1 : : : zn t1 : : : tq (2. These are the conditionally Gaussian model. There are various models of the type (2.12) which consists of the observed factors and the components of the lter state process.14) t t t 5 . This means.11) ~t which is an Ak -adapted process with a certain dimension q 1. We assume that the SDEs (2. we assume that the lter state t satis es an SDE of the form d t = Ct (~tk ) dt + Dt. Example 2.13).10) for all t 2 0 T ].3) takes the form dxt = a0 + a1 xt + a2 yt dt + bt dwt t t t dyt = A0 + A1 xt + A2 yt dt + Bt (yt) dvt (2. ztk = yt1 : : : ytk t1 : : : tq > ~ (2. Furthermore.1 : Conditionally Gaussian Filter Model Assume that in the system of SDEs (2. while gt(zt ) Gt(yt ) 0. characterized by the lter state process k t n t = n t= . This means the model (2. that we have a nite-dimensional lter.lter and the nite-state jump model for x. t 1 ::: q > t t 2 0 T] o (2. which leads to a generalized Kalman. t 2 0 T ].latter characterizes completely the entire lter distribution. considered in Liptser & Shiryaev (1977). ) dyt z z (2. Considering conditional expectations of integrable functions of xt is thus not a restriction for the identi cation of lter equations.3) that admit a nite-dimensional lter with t satisfying an equation of the form (2. These are. We shall denote by ztk the resulting (k + q)-vector of observables ~ . The general case of lter equations is beyond the scope of this paper. for instance.3) are such that the corresponding lter distributions admit a representation of the form .(~tk.13) with Ct( ) denoting a q-vector valued function and Dt ( ) a (q k)-matrix valued function. which is related to hidden Markov chain lters. as demonstrated in the literature that we mentioned in the introduction.3) the functions at ( ) and At ( ) are linear in the factors and that bt (zt ) bt is a deterministic function. Various combinations of these models have nite-dimensional lters and can be readily applied in nance.

this is su cient to determine the solution of (2. t t . k) and k respectively. A1 .k) > ) . The value of Bt(yt) becomes known only at time t. then we have a Gaussian lter model. o > + ct (A1 )> . t t t t Bt (yt) are matrices of appropriate dimensions. where the lter process is given by the vector process = f t t 2 0 T ]g and the upper triangular array of the elements of the matrix process c = fct t 2 0 T ]g with q = (n . A2 . then ct can be computed o -line. Model (2. In this case the lter distribution is a Gaussian distribution with vector mean (n. Although for t 2 0 T ].14) is in fact of the type of a conditionally Gaussian lter model. in the case when Bt (yt) does not depend on yt for all t 2 0 T ]. Recall that w is an n-dimensional (A P )-Wiener process and v the vector of its rst k components.17) where bt is the k-dimensional vector obtained from the rst k components of bt . 0 dyt .17).17) that if Bt (yt) does not depend on the observable factors yt. Obviously. and a1 . a2.1 bt Bt (yt )> + ct (A1 )> > dt bt Bt (yt) t t (2.kg.xi . 2 Note by (2. At + A1 t + A2 yt dt t t dct = a1 ct + ct (a1)> + (bt b>) t t t . namely d t = a0 + a1 t + a2 yt dt + bt Bt (yt)> + ct (A1 )> (Bt(yt) Bt (yt)>). except for Bt (yt) that depends also on observed factors. 6 . Notice that the computation of ct is contingent upon the knowledge of the coe cients in the second equation of (2.for t 2 0 T ] with given deterministic initial values x0 and y0. bt .16) that the matrix ct is symmetric.17) at time t. it follows from (2. The above lter can be obtained from a generalization of the well-known Kalman lter.k)] . t 2 0 T ]. however. ct is de ned as a conditional expectation.Bt (yt ) Bt (yt )> . These coe cients are given deterministic functions of time. i t t t (2. where t c` i = E x` . see Chapter 10 in Liptser & Shiryaev (1977).16) The dependence of t and ct on k is for simplicity suppressed in our notation. (2.15) ~t Ak : and covariance matrix ct = c` i]` i2f1 2 ::: n. Here a0 and A0 are t t column vectors of dimensions (n . where t = ( 1 ::: t t i t ~t = E xit Ak ` . k) 3+(n. We recall that Bt (yt) is assumed to be invertible.1 t t t t .

we consider the particular case of model (2.k )> t 2 0 T ]g.3) with Gt(yt ) 0. At(yt pt) dt i (2.2 : Finite-State Jump Model Here we assume that the unobserved factors form a continuous time. k)dimensional jump process x = fxt = (x1 : : : xn. Let ait j (y h) denote the transition kernel for x at time t to jump from state i into state j given yt = y and xt = h.(zt.19) j 2 f1 2 : : : M g. and an intensity vector t(zt ) = ( 1(zt ) : : : n(zt ))> t t at time t 2 0 T ] for the vector counting process N = fNt = (Nt1 : : : Ntn)>. The vector pt satis es the following dynamics h i j = . Liptser & Shiryaev (1977). Chapter 9. the lter distribution is completely characterized by the vector of conditional probabilities pt = (p1 : : : pM )>. More precisely.21) 7 .Example 2. which can take t t a nite number M of values.18) for t 2 0 T ]. for the vector yt of observed factors we assume that it satis es the second equation in (2. On the other hand. In this example.B (y ) B (y )> . at (yt pt )> pt j ~ = M X i=1 ait j (yt h) ph t x= t j pit At (yt j ) = At (yt xt ) ~ At (yt pt) = M X j =1 At(yt j ) pj t (2. Notice that the process x of unobserved factors has here only jumps and is therefore piecewise constant. where in the xt dynamics we have at (zt ) = gt(zt ) t (zt ) and bt (zt ) 0. by (2. (n . where M is the number of possible t t states 1 : : : M of the vector xt and pj = P xt = t j ~t Ak (2. This means that the process of observed factors y is only perturbed by continuous noise and does not jump. A (y p ) .20) M X h=1 ! see. given an appropriate time t and zt -dependent matrix gt (zt ).3) and (2. Thus.3).7) we have dxt = gt. ) dNt (2.1 ~t t t dpt ~t t t t t t t t t t t h ~ dyt . t 2 0 T ]g.a (y p )> p j dt + pj A (y j ) . where .

we obtain the representation ~ z ~t At (~tk ) = E At(zt ) Ak (2.s.24) for t 2 0 T ].3).25) ~ P -a.3 that Let At (zt ) and the invertible matrix Bt(yt ) in (2.3) be such Z T 0 Z T 0 E (jAt (zt )j) dt < 1 and Bt (yt) Bt(yt )> dt < 1 (2.12). j 2 f1 2 : : : M g.) dNt ~ 8 (2. ~ z the expression At(~tk ) takes the particular form ~ z At (~tk ) = A0 + A1 t + A2 yt: (2. 1. (yt. t 2 0 T ]g with q = M .23) t t t ~ z Furthermore.2. Since the lter is characterized by the lter state process . ~ z dyt = At (~tk ) dt + Bt (yt) dvt + Gt. 2. The lter state process = f t = ( t1 : : : tq )>. for Example 2.for t 2 0 T ]. Then there exists a k-dimensional Ak -adapted Wiener process v = fvt t 2 ~ ~ 0 T ]g such that the process y = fyt t 2 0 T ]g of observed factors in (2. see (2. which provides an important representation of the SDE for the observed factors. given in (2. namely the conditionally Gaussian model. t 2 0 T ]g for the model of this example is thus given by the vector process p = fpt = (p1 : : : pq )>. we need only M . in general. Since the probabilities add t t to one.22) where the vector ztk is as de ned in (2.k ).26) . This is the conditional 1 expectation of At (zt ) = At (yt : : : ytk x1 : : : xtn.12 in Liptser & Shiryaev (1977).2 Markovian Representation As in the two previous examples we have.22). with respect t to the lter distribution at time t for the unobserved factors xt .1. At(~tk ) can be represented as ~ z ~ At (~tk ) = At(yt pt) = M X j =1 At(yt j ) pj t (2. in our lter setup to deal ~t with the quantity E (At (zt ) j Ak) assuming that it exists.21). in the case of Example 2. We have now the following generalization of Theorem 7. 1 probabilities to compute. Proposition 2. ~ Notice that. namely the nite-state jump model.3) satis es the SDE ~ z with At (~tk ) as in (2.

26) and (2.26).28).27) ~ z ~ z ~ z whereby we implicitly de ne the vector Ct(~tk ) and the matrices Dt(~tk ) and Gt (~tk ) for compact notation. G. From equations (2. The functions.(yt. D and G appearing in (2.3). Corollary 2. as well ~ as property (2.27) we immediately obtain the following result. zt1.(~tk. (~tk. t t where xt = (x1 : : : xn. Instead of the original factors zt = (yt1 : : : ytk x1 : : : xn. we may now base our analysis on the t t components of the vector ztk = (yt1 : : : ytk t1 : : : tq )>. ) Gt.k )> = (zt1 : : : ztn )>. that are all ~ observed. projected into a Markovian model for the ~ observed quantities. see (2. q dNtr for t 2 0 T ] and ` 2 f1 2 : : : k + qg. replacing dyt in (2.10).k )> is unobserved.(~tk.29) + k X r=1 `r + t.) dNt z z ~ z z ~ z ~ z ~ z ~ ~ z = Ct (~tk ) dt + Dt(~tk ) dvt + Gt.12).13) ~ ~ by its expression resulting from (2. we obtain h i d t = Ct(~tk ) + Dt(~tk ) At(~tk ) dt + Dt (~tk ) Bt(yt) dvt + Dt. C . ~ ~ Due to the existence of a Markovian lter dynamics we have our original Markovian factor model.(yt.28) From Corollary 2. In fact. Here the driving observable noise v is an (Ak P )-Wiener ~ process and the observable counting process N is generated by the rst k components N 1 N 2 : : : N k of the n counting processes.3 is given in Appendix A. ) dNt: (2. Just as was the case with z = fzt t 2 0 T ]g. : : : ztk. also the vector process zk = fztk t 2 0 T ]g has a Markovian dynamics. `. zt2.The proof of Proposition 2. ` r and ` r follow ~ ~ ~ ~ directly from A. ) dNt (2. B . For e cient notation we write for the vector of observables ztk = zt = (zt1 zt2 ~ : : : ztk+q )> the corresponding system of SDEs in the form dzt` = ` (t zt zt : : : 1 2 k k+q ) dt + X ` r (t z 1 z 2 zt t t r=1 : : : ztk+q ) dvtr ~ (2.4 The dynamics of the vector ztk = (yt t ) can be expressed by ~ the system of SDEs ~ z dyt = At (~tk ) dt + Bt(yt) dvt + Gt. 9 .4 it follows that the process zk = fztk t 2 0 T ]g is Markov. the following result. given by (2. We also have as an immediate consequence of the Markovianity of zk = z .) dNt ~ ~ z ~ z ~ ~ z d t = Ct(~tk ) dt + Dt (~tk ) dvt + Gt.

for instance. As described in Platen (2002). We then denote by S j (t) the discounted value at time t of the j th primary security account.1 Primary Security Accounts We assume that there are d+1 primary assets in the market. 3.30) Relation (2.5 ~t Any expectation of the form E (u(t zt ) j Ak ) < 1 for a given function u : 0 T ] <n ! < and given k 2 f1 2 : : : n . see Long (1990). We call B 0 also the 0th primary security account process. For instance. We assume that S j is Ak -adapted and the 10 . ~ (2. 1g can be expressed as ~t E u(t zt ) Ak = uk (t ztk ) ~ ~ with a suitable function uk : 0 T ] <k+q ! <. we model the di erent denominations of the growth optimal portfolio (GOP). where d = 2 k.30) in Corollary 2. We only use the observed factors to model the GOP. These are. currencies or shares. generated by the observed factors introduced above. S j (t) is the value of the savings account of the j th foreign currency. For the domestic currency as primary asset we express the time evolution of its value by the savings account process B 0 = fB 0(t) t 2 0 T ]g. j 2 f1 2 : : : dg. then S j (t) is the cum-dividend share price.5 will be of importance for contingent claim pricing as we shall see later on.1) ~ for t 2 0 T ] and j 2 f0 1 : : : dg. where all dividend payments are reinvested.Corollary 2. we formulate a ltered benchmark model. these factors evolve in conjunction with unobserved factors that in uence the observed ones. expressed in units of the domestic currency. in the case of currencies. However. For the modelling of the time value of the j th primary asset. The resulting ltered benchmark model has the key advantage that a consistent price system is automatically established without using any measure transformation. that is Sj S j (t) = B 0(tt) () (3. 3 Filtered Benchmark Model On the basis of the Markovian dynamics for the prices. If the j th asset is a share. we introduce the j th primary security account process S j = fS j (t) t 2 0 T ]g.

no primary security account can be expressed as a self.) + k X r=1 'j r (t. 3. ~ We assume that j r and 'j r are Ak -predictable with 'j r (t) 0. The given parameterization of the above SDE (3.unique strong solution of the stochastic di erential equation (SDE) dS j (t) = S j (t.'0 r (t) ~rt (yt) dt + dNtr for t 2 0 T ] with S j (0) > 0. j r (t)) ( 0r (t) dt + dvtr ) ~ (3.) '0 r (t.nancing portfolio of other pri11 .) ( 0r (t) .) dS j (t) (3. t 2 0 T ]g is a predictable stochastic process = f (t) = ( (t) : : : self. That means. Under a self. the corresponding portfolio V 0(t) has at time t the discounted value d V 0 (t) = X j (t) S j (t) V (t) = B 0(t) j =0 0 (3.nancing strategy expresses the number of units of the j th primary security account held at time t in the corresponding portfolio. '0 r (t) > 0 a.2) .4) for all t 2 0 T ].2 Portfolios ~ Let us now form portfolios of primary security accounts. if is S -integrable.nancing strategy. j 2 f1 2 : : : dg. 1 . We assume that no primary security account is redundant. We say that an Ak 0 d (t))>.3) and it is dV 0 (t) = d X j =0 j (t.nancing strategy no out ow or in ow of funds occurs for the corresponding portfolio. for t 2 0 T ] and Z T 0 ( j r (s))2 + ~ r (y ) s s ds < 1 for j 2 f1 2 : : : dg and r 2 f1 2 : : : kg.s. The j th component j (t). j 2 f0 1 : : : dg. All changes in the value of the portfolio are due to gains from trade in the primary security accounts.2) does not restrict its generality but is convenient for the benchmark approach. of the self. see Protter (1990).

we specify V i(t) as a function of time t and the vector of observables ztk .3 Growth Optimal Portfolio The GOP is the self.1 'i r (t. Assuming su cient smoothness of V i( ).mary security accounts. j r (t) q for r 2 f1 2 : : : kg ' . Let us set bj r (t) = 8 > < > : 0r (t) . Note that the observed market is complete. We denote by V i (t) the value of the GOP when it is expressed at time t in units of the ith primary security account. which means that the observed primary security accounts securitize the uncertainty generated by the Wiener processes v1 : : : vk and the counting processes N 1 : : : N k .5) for t 2 0 T ] and j 2 f1 2 : : : dg.1 ~r.k (yt) for r 2 fk + 1 : : : dg t (3. It has the form dV i(t) = V i (t.) dNsr ~k 12 .) (3.6) . For the di usion case without jumps the corresponding SDE is well known.29). we then obtain o V i(t ztk ) ~ = V i(0 zk ) + ~ 0 Z t 0 L 0 V i(s zs ) ds + ~k k XZ t r=1 0 Lr V i(s zs ) dvs ~k ~r (3. In the case with jumps the derivation of the SDE for the GOP is more involved and described in Platen (2002). that is ~ ~ V i(t) = V i (t ztk ) = V i(t zt1 : : : ztk+q ) (3. given our factor model. ~ ~ 3. (t) '0 . We then de ne the matrix b(t) = bj r (t)]d r=1 j for t 2 0 T ] and assume that b(t) is for Lebesgue-almost-every t 2 0 T ] invertible.'i r (t) ~ rt (yt) dt + dNtr for t 2 0 T ] and i 2 f0 1 : : : dg. Long (1990) or Karatzas & Shreve (1998). for instance. by application of the It^ formula and using (2.nancing portfolio that achieves maximum expected logarithmic utility. To make the above framework computationally tractable.8) + k XZ t r=1 0 r V i (s.7) for t 2 0 T ] and i 2 f0 1 : : : dg. (t) j r k r k . zs.) k X r=1 i r (t) ( i r (t) + dv r (t)) + ~ o 1 . see.

~t.for t 2 0 T ] and i 2 f0 1 : : : dg. q + . q : : : ztk.) = ' (3. zt1. Here we use the operators k+q @ X L0 = @t + `=1 ` t zt1 : : : ztk+q @@ ` z `r +1 2 k+q k XX ` p=1 r=1 t zt1 : : : ztk+q = k+q X `=1 `r pr t zt1 : : : ztk+q @ z@@ zp ` 2 (3. see (3. q + k+q r + t.6) and (3. F t. By comparison of (3. F t zt1 : : : ztk+q @@ ` z F t zt1. This means for j 2 f0 1 : : : dg that the j th benchmarked primary 13 . z k ) + V i (t.6) is given by the expression V i(t. r 2 f1 2 : : : kg and F : 0 T ] <k+q ! < being any given function of time and vector of observables.9) (3.8) it follows that the i `th GOP-volatility i `(t) has the form ` i ~k i `(t) = L V (t zt ) (3.12) ~ V i(t ztk ) and the inverted i `th GOP-jump ratio 'i `(t).13) ` (t. ztk. where ztk = zt = (zt1 : : : ztk+q )> as ~ introduced before (2. benchmarked prices.29). zt1. ) ~ i ` (t. : : : ztk.11) for t 2 0 T ]. : : : ztk. z k ) = ~t. z k ) ~t. : : : ztk. i 2 f0 1 : : : dg. q (3. 4 Fair Pricing and Hedging of Derivatives 4.10) Lr and r (t. + 1r + + t. ` 2 f1 2 : : : kg. zt1. V i for t 2 0 T ].1 Benchmarked Prices In what follows we call prices that are expressed in units of the GOP.

(3. it is impossible to generate. Then all prices. in general. d X j =0 j (t) j r (t) dv r (t) ~ ! d X j =0 j (t. We underline that such assumptions will not be needed for our results. Note that ~k P )-martingales. j r (t) dvtr + ('j r (t. 1) dmrt ~ (4.) .2). strictly positive wealth from zero initial capital.6) and application of the It^ formula the SDE o ^ ^ dS j (t) = S j (t. 4. (3. discounted by the domestic savings ~ account B 0 would be (Ak P k)-martingales. let us discuss a situation where we assume for the moment that the following steps can be made and a standard equivalent risk neutral probability measure P k exists.2 Derivative Prices To provide an intuitive link between the benchmark framework and the standard risk neutral approach.5). with strictly positive probability. Moreover.) 'j r (t.) k X r=1 .) + k X r=1 ( . see Platen (2002). Note that the j th bench~ marked primary security account is an (Ak P )-local martingale. not (A ~ benchmarked portfolio process is here an (Ak P )-supermartingale. Here mr denotes the rth t component of the jump martingale m de ned in (2.nancing portfolio V 0 it 0follows 0by applicaV V ^ tion of the It^ formula that its benchmarked value V (t) = V 0((tt)) = V 0 ((tt)) satis es o the SDE ^ ^ dV (t) = V (t. Since a nonnegative these processes are.7). Denoting by E the expectation with respect to P and by E k that with respect to P k .2) for t 2 0 T ] and j 2 f0 1 : : : dg. as shown in Platen (2002).^ ^ security account S j = fS j (t) t 2 0 T ]g with j Sj ^ S j (t) = V 0(tt) = S 0(t) ( ) V (t) (4. we would have.1) satis es by (3. 1 dmr t ) (4.3) ^ ~ for t 2 0 T ]. the resulting ltered benchmark model can be shown to exclude standard arbitrage. taking into 14 . This shows that V is an (Ak P )-local martingale too.) . for any self. This means.

7) for all t 2 0 ]. A price process is called fair. if its benchmarked values form an ~ (Ak P )-martingale.5) for t 2 0 T ].nancing portfolio V 0 would satisfy the relation B 0(t) V 0( ) Ak ~t V (t) = B0( ) k B0( ~t = E k B 0 ( t) V 0 ( ) Ak ) t ! V 0( ) Ak ~ = V 0(t) E 0 V () t 0 Ek (4. Furthermore.6) ~ for t 2 0 ] and any Ak -stopping time 0 . Here the Radon-Nikodym derivative would satisfy the expression 0 0 ^0 k = V (0) B (t) = S (t) t ^ V 0(t) B 0(0) S 0(0) (4. by (4. we consider a benchmarked contingent claim U ( y ) as a function of and the corresponding values of observed factors y . which is assumed to be an Ak -stopping time. under the above assumptions all V ~ ^ benchmarked portfolio prices V (t) = V 0 ((tt)) would be (Ak P )-martingales. In the benchmark framework we avoid the above steps and the assumption on the existence of an equivalent risk neutral measure by introducing the concept of a fair price.1).5). ~ At a given maturity date .account (4. that is ^ ^ ~t V (t) = E V ( ) Ak (4.4) = dP dP k for t 2 0 T ] and j 2 f1 2 : : : dg. Thus. the price of a self. see Platen (2002).8) 15 . where we assume that ~t E (jU ( y )j Ak ) < 1 (4. that S j (t) B 0 (t) S j (T ) Ak ~t = B 0 (T ) k 0 ~t = E T B0 (t) S j (T ) Ak k B (T ) t ! j (T ) S ~t = V 0 (t) E V 0(T ) Ak Ek k T (4.

12) is that it allows us to express the benchmarked ~t fair price uk (t ztk ) as conditional expectation with respect to Ak . for all t 2 0 ].11) for t 2 0 ]. The corresponding fair price at time t for this contingent claim. The above derivation can be summarized in the following result. The actual ~ ~ computation of the conditional expectation in (4.1 The benchmarked fair price uk (t ztk ) for the benchmarked con~ ~ tingent claim U ( y ) can be expressed as the conditional expectation ~t uk (t ztk ) = E u(t zt) Ak ~ ~ for t 2 0 ]. The advantage of the representation (4. (4.1 the information. Corollary 4.1. This means. This al~ lows us to de ne the (A P )-martingale u = fu(t zt ) t 2 0 ]g by the conditional expectation . when expressed in units of the domestic currency. obtained by the conditional expectation ~t uk (t ztk ) = E U ( y ) Ak ~ ~ (4.10) for t 2 0 ).12) ~t We recall that Ak denotes in Corollary 4. otherwise the payo would not be veri able at time .s. we form directly the conditional expectation (4. The benchmarked fair price process uk = fuk (t ztk ) t 2 0 ]g for the ~ ~ ~ ~ benchmarked contingent claim U ( y ) is then the (Ak P )-martingale. is then u0 k (t ztk ) = V 0 (t) uk(t ztk ) ~ ~ ~ ~ (4. which is available at time t. not only of zk but also of z .30). Note that the benchmarked fair price. given in Corollary 4. ts perfectly the expression of our result for the ltered factor model given in (2. The above concept of fair pricing generalizes the well-known concept of risk neutral pricing and avoids not only the assumption on the existence of an equivalent risk neutral measure but also some issues that arise from measure changes under di erent ltrations.9) for t 2 0 ].12) is equivalent to the solution of the ltering problem for the unobserved factors. There is no point to let the payo function depend on any other than observed factors.7) without using any measure transformation. whereas At is the complete information at time t that determines the original model dynamics including also the unobserved factors. u(t zt) = E U ( y ) At (4. The vector of observables y is a subvector.a. which at time t exploits the complete information characterized by the -algebra At. 16 .

we consider ~ ^ V (t) with U U ^ dV (t) = U d X j =0 j ^j U (t.12) ~ is di erentiable with respect to time and twice di erentiable with respect to the observables.12) and (3. To replicate the benchmarked contingent claim U ( y ) we can start at a given time t 2 0 ] by forming a portfolio with fair benchmarked price ^ ~t V (t) = uk (t ztk ) = E U ( y ) Ak : ~ ~ (4. Let us search for a fair benchmarked price process V .18) ~ ~t j =0 U 17 . Above we used the j th proportion j ^j j (t) = U (t) S (t) (4.3 Hedging Strategy Assume that the above benchmarked pricing function uk ( ) in (4.) dS (t) (4.) . By (4. that possibly matches uk . z k ) ~ uk (s.) uu (s.3) we then have ^ ^ V ( ) = V (t) .15) the proportions must satisfy the system of linear equations d L` uk (t ztk ) = X j (t) j `(t) ~ ~ .4.14) for t 2 0 ]. U U k XZ `=1 t U ^ V (s) U d X j =0 j U j U (s) j `(s) dv ` ~s ! + k XZ `=1 t ^ V (s.17) U U U By comparison of (4.15) Note that the volatilities and jump ratios in (4.13). with selfnancing hedging strategy U . uk (t zk ) (4.9) and (4. 1 dm` : s (4. Then we obtain by the It^ formula the representation o U ( y ) = uk ( z k ) ~ ~ = uk (t ztk ) + ~ ~ + k XZ k XZ `=1 t ` ~k k uk (s zs ) Luk (s(szkz)s ) dvs ~ ~k u ~ ~` ~ ~ s k ` (s. zs. dm` ~ ~k ~ (4.13) and (4.)'j r (s.15) are those identi ed in (3. zks.) d X j =0 (s. This means.16) ^ V (t) of the value of the corresponding hedging portfolio that has to be invested into the j th primary security account at time t 2 0 ].13) ~k ~s.) s t `=1 ^ for t 2 0 ].

) (4. In such a case the market is incomplete. z k ) ~t.)> b.21) ~ for t 2 0 T ]. we obtain the following result. Let us use the d-dimensional vector c(t.) for r 2 fk + 1 : : : dg t (4.(4.19) for ` 2 f1 2 : : : kg and t 2 0 T ].20) > : '0 `1 t.2 For a given benchmarked contingent claim U ( y ) with corresponding vector cu (t) given in (4.) = cu (t. u ~ uk (t.12) on the basis of the lter distribution. ) ~ ~ u (t.20) the proportions of the corresponding hedg~ ing portfolio are of the form k U .1.and ` (t. ) ~ ~ r k k t k k t + 0r (t) q for r 2 f1 2 : : : kg ~r.1 t 2 0 ).22) for t 2 0 ). k~t.5). k U Proposition 4.(2. ) ~ ~ r uk k +1 .19) in the form cu (t.2) . z . ) ~ ~ k +1= d X j =0 j (t.k (yt. By involving the matrix b(t) given in (3. the degree of available information is indexed 18 . see (2.)> = (t. we can then rewrite the system of equations (4.) : : : cd(t. (t. zt. ztk.18) .) 'j `(t.) (4.4 Variance of Benchmarked Prices Let us now investigate the impact of varying degrees of information concerning the factors zt = (zt1 : : : ztn)> that underly our model dynamics. z .) ( . Note that the invertibility of the matrix b(t) is not linked to a speci c contingent claim. Incomplete markets of this type can be handled by a generalization of the above described ltered benchmark approach. The benchmarked pricing functions can always be obtained from the conditional expectation (4.)> b(t. Now.1 (t. We did not consider the case d < 2k. 4.) = (c1(t.))> with components cr~ u k (t.3). one can form a perfectly replicating hedging portfolio for all benchmarked contingent claims U ( y ).) c2 (t. The introduced ltered benchmark model forms a complete market despite the fact that the original model involves unobserved factors.) > ~ k (4.k z k ~ (t. As already mentioned in Section 2. Thus.) = > 8 > > < L u (t. ) u (t.

27) 19 . uk (t ztk ) 2 = u(t zt) . uk (t ztk ) ~ ~ ~ ~ . uk (t ztk ) : ~ ~ ~ ~ ~ ~ (4. which expresses the reduction in conditional variance and can also be seen as a generalization of the celebrated RaoBlackwell theorem towards ltering. uk (t ztk ) 2 Ak ~ ~ (4.26) where for t 2 0 ). For k 2 fr r + 1 : : : n . Rtk+m t t (4. . 1g we have . Note that for larger k we have more information available. uk (t ztk ) 2 ~ ~ ~ ~ ~ ~ ~ ~ . Proposition 4. A larger value of k means that more factors are observed. ~ Consider from now on a benchmarked contingent claim U ( y ) = U ( y1 y2 : : : yr ) (4. Rtk+m = E uk+m(t ztk+m ) .12). kg and k 2 fr r + 1 : : : n . 1g let ~t uk (t ztk ) be the corresponding benchmarked fair price under the information Ak . 2 ~t Ak For t 2 0 ) and k 2 fr r + 1 : : : n . This is avoided by using the suggested ltered benchmark model. which naturally should reduce the conditional variance. For each degree of available information one obtains. ~ ~ as given by (4. Proof: . uk+m(t ztk+m ) u (t ztk+m ) . the process zk is Markovian. . k+m + 2 u(t zt) . We can prove the following proposition.28). Then . 1g we ~t E Vark+m(u) Ak = Vark (u) .12). ~t Vark (u) = E u(t zt ) .23) for some xed r 2 f1 2 : : : n . Again we use the notation ztk for the ~ vector of observables de ned in (2. uk+m(t ztk+m ) 2 + uk+m(t ztk+m ) . u(t zt ) . Furthermore.by the parameter k. in general.3 have For m 2 f0 1 : : : n . Recall by (2. where we stress its dependence on k and recall that. All conditional expectations can be taken under the real world probability measure P .30) that uk (t ztk ) can be computed as conditional ~ ~ expectation via the lter distribution.25) (4. di erent equivalent risk neutral probability measures. where we assume that the number of observed factors that in uence the claim equals r.24) t is the corresponding conditional variance at time t 2 0 ). 1g. ~ providing thus more information in Ak . The complexity of working with di erent pricing measures can be signi cant. it is clear that ltering itself is always performed under the real world measure. by (2.

(A. we obtain (4. that is ytc = yt .(y .3) . Let us now de ne the k-dimensional ~ Ak -adapted process v = fvt t 2 0 T ]g by ~ ~ ~ z Bt (yt) dvt = dytc . The reduction of the conditional variance of fair derivative prices under increased information is quanti ed via a generalization of the Rao-Blackwell theorem. At (~tk ) dt: ~ (A. benchmarked security prices are not forced to be martingales. 5 Conclusions We constructed a ltered benchmark model by specifying the growth optimal portfolio for a given degree of available information. They may be just local martingales. is the vector ( N 1 . X j t G .2) j j j From (2. A Appendix Proof of Proposition 2.25). uk+m(t ztk+m) Ak+m ~ ~ ~t Ak : (4. : : : N k .3 Denote by yc the continuous part of the observation process y.)>.1) and (A.1 At (zt ) . A consistent price system has been established.28) Since the last term on the right hand side is equal to zero by de nition.2) it follows that ~ z dvt = dvt + Bt(yt ).) N j j j (A. N .~t By taking conditional expectations with respect to Ak on both sides of the above equation it follows that . ~t Vark (u) = E Vark+m(u) Ak + Rtk+m + 2 E uk+m(t ztk+m ) . uk (t ztk ) ~ ~ ~ ~ t t .1) j j where the j denote the jump times of N = fNt t 2 0 T ]g and N = N . In general. ~t E u(t zt) . At (~tk ) dt: ~ 20 h i (A.3). Benchmarked fair derivative prices are obtained as martingales under the real world probability measure.

2 v ~ which has the solution 0Z t s s exp { (~u . Chiarella. vs) is a k-dimensional vector v ~ ~k -measurable Gaussian random variables. 2 exp { (~u . v ~ s s ~s exp { (~u . ~s E exp { (~u .4) we end up with the equation ~s E exp ({ (~t . Au(~u ) ~ ~s Au du Ak = 0: (A. by our assumptions and by the boundedness of exp { (~u . vs)] Ak = exp . vs)] du: v ~ s ~ Recalling that v is an Ak -measurable Wiener process. vs)] Ak du (A. Estimation in models of the instantaneous short term interest rate by use of a dynamic Baysian algorithm. vs)] dvu Ak = 0 v ~ (A. Runggaldier (2001a). that exp { (~t . s) v ~ for 0 s t T . vs)]. Au(~u ) du v ~ . (A. Sydney. QFRG Research Paper 68. ~ zk Au(zu) . s) and independent of Ak . We can conclude that (~t . A References Bhar. 2 (t . v is thus a k-dimensional ~ ~k -adapted standard Wiener process. vs)] Bu 1 (yu) Au(zu ) . University of Technology. vs)] Bu 1(yu) Au(zu) . & W. 21 .5) ~ zk ~s exp { (~u . vs)]) Ak = 1 .6) Taking conditional expectations on the left and the right hand sides of (A. C. Technical report.4) and that. Au(~u ) du Ak = v ~ . notice that E E E Z t Z t Z t > Z t s ~ zk exp { (~u . by the multi-variate It^ formula with 2 <k a row vector and o { the imaginary unit. R.8) E exp { (~t ..7) v ~ > ~s (A. By Levy's theorem. vs)] = 1 + { v ~ +{ Z t Z t s exp { (~u .From this we nd. each with variance of independent At ~s (t . vs)] Bu 1(yu)E v ~ . vs)] dvu v ~ .

In Seminar on Stochastic Analysis. University of Technology. Appl. S. Finance and Stochastics 4. Sydney. I. Stochastic Integration and Di erential Equations. 33(4). Platen (2001). C. QFRG Research Paper 69. R. E. Filtering equity risk premia from derivative prices. Frey. 29{69. Adv. Appl. Gombani. In Stochastics in Finite and In nite Dimensions. J. B. Statistics of Random Processes: I. Fischer. Runggaldier (2001). in Appl. J. Fischer. Technical report. Random Fields and Applications. Platen. Theor.. J. Karatzas. 794{809. R. Monte Carlo Methods Appl. 229{238. Finance 4(2). Birkhauser. Runggaldier (2001b).. General Theory. A ltering approach to pricing in multifactor term structure models. & W. R. J.. A nonlinear ltering approach to volatility estimation with a view towards high frequency data. Elliott. R. Fischer. Int. 371{385. Probab. & E. Long. C. Chiarella. Volume 39 of Appl. Landen. 175{ 188. Finance and Stochastics 1. J. Finance 4(2). Runggaldier (2001). IEEE Control Systems Society. Runggaldier (2001). & J. University of Technology Sydney. Liptser. Benchmark models with intensity based jumps.Bhar. P. Probab. & W. Elliott. (2000). Shiryaev (1977). & W. 303{ 320. An application of hidden Markov models to asset allocation problems. Hidden Markov chain ltering for generalised Bessel processes. Math. Int. (working paper). Financial Economics 26. pp. Platen. Springer. Theor. (1990). & W.. Hidden Markov ltering for a mean reverting interest rate model. E. & A. Risk minimizing hedging strategies under partial observation. P. Platen (1999). P. 19{ 38. & S. A. 22 .. 123{ 143. pp. & W. Elliott. & E. Volume 45 of Progr. pp. R. Volume 5 of Appl. Bond pricing in a hidden Markov model of the short rate. Springer. Applications of the balanced method to stochastic di erential equations in ltering. Platen (1999). In 38th IEEE CDC meeting in Phoenix. Springer. J. Shreve (1998). J. J. Pasquali. 2782{2787. 199{210. E. van der Hoek (1997). C. Methods of Mathematical Finance. & E. (1990). R. Chiarella. J. 5(1). Birkhauser. Protter. P. On ltering in Markovian term structure models (an approximation approach). (2002). The numeraire portfolio. Runggaldier (1999). Math.

- Black Litterman Idzorek.pdf
- Portfolio Risk Analysis of 5 stocks
- Erc Slides
- 9-29-2015_Stochastic
- 2010_Fabozzi y Haung_Robust Portfolio Contributions From
- Chapter 10 and 11
- SAPM Protfollio Section
- sapm
- francesco
- Risk Aversion
- Optimal Mean-Variance Investment Strategy Under
- Finance Summaries
- 05+-+CAPM
- Goldman
- 301 AssetPricing v 07 4 Per Page
- FAC
- America Latina
- capm
- Risk & Return
- Calculation of HPY and Risks of Investment
- Risk and Return
- Dhaka Stock Exchange Investment Simu
- SSRN-id297639
- execution risk
- BarraOne for Asset Managers April 2010
- ch5_inatlbond.pdf
- Algo Risk Service
- capm
- Portfolio Selection
- Capital Asset Pricing Model

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue reading from where you left off, or restart the preview.

scribd