Optimal Portfolio under Fractional Stochastic Environment

Similar documents
AMartingaleApproachforFractionalBrownian Motions and Related Path Dependent PDEs

6. Stochastic calculus with jump processes

An Introduction to Backward Stochastic Differential Equations (BSDEs) PIMS Summer School 2016 in Mathematical Finance.

f(s)dw Solution 1. Approximate f by piece-wise constant left-continuous non-random functions f n such that (f(s) f n (s)) 2 ds 0.

An Introduction to Malliavin calculus and its applications

Utility maximization in incomplete markets

Risk Aversion Asymptotics for Power Utility Maximization

Stochastic Modelling in Finance - Solutions to sheet 8

Optimal Investment, Consumption and Retirement Decision with Disutility and Borrowing Constraints

IMA Preprint Series # 2056

CH Sean Han QF, NTHU, Taiwan BFS2010. (Joint work with T.-Y. Chen and W.-H. Liu)

Optimal Investment Strategy Insurance Company

Quadratic and Superquadratic BSDEs and Related PDEs

Uniqueness of solutions to quadratic BSDEs. BSDEs with convex generators and unbounded terminal conditions

Lecture #31, 32: The Ornstein-Uhlenbeck Process as a Model of Volatility

Generalized Snell envelope and BSDE With Two general Reflecting Barriers

Optimal Investment under Dynamic Risk Constraints and Partial Information

A general continuous auction system in presence of insiders

Backward stochastic dynamics on a filtered probability space

Portfolio optimization for a large investor under partial information and price impact

Homework 10 (Stats 620, Winter 2017) Due Tuesday April 18, in class Questions are derived from problems in Stochastic Processes by S. Ross.

Chapter 14 Wiener Processes and Itô s Lemma. Options, Futures, and Other Derivatives, 9th Edition, Copyright John C. Hull

and Applications Alexander Steinicke University of Graz Vienna Seminar in Mathematical Finance and Probability,

Existence and uniqueness of solution for multidimensional BSDE with local conditions on the coefficient

Some new results on homothetic forward performance processes

FINM 6900 Finance Theory

Physics 127b: Statistical Mechanics. Fokker-Planck Equation. Time Evolution

Algorithmic Trading: Optimal Control PIMS Summer School

On a Fractional Stochastic Landau-Ginzburg Equation

What Ties Return Volatilities to Price Valuations and Fundamentals? On-Line Appendix

School and Workshop on Market Microstructure: Design, Efficiency and Statistical Regularities March 2011

OBJECTIVES OF TIME SERIES ANALYSIS

Sample Autocorrelations for Financial Time Series Models. Richard A. Davis Colorado State University Thomas Mikosch University of Copenhagen

1. Consider a pure-exchange economy with stochastic endowments. The state of the economy

MATH 5720: Gradient Methods Hung Phan, UMass Lowell October 4, 2018

Option pricing and implied volatilities in a 2-hypergeometric stochastic volatility model

Richard A. Davis Colorado State University Bojan Basrak Eurandom Thomas Mikosch University of Groningen

Vanishing Viscosity Method. There are another instructive and perhaps more natural discontinuous solutions of the conservation law

14 Autoregressive Moving Average Models

Stochastic Model for Cancer Cell Growth through Single Forward Mutation

Cash Flow Valuation Mode Lin Discrete Time

1 THE MODEL. Monique PONTIER U.M.R. CNRS C 5583, L.S.P. Université Paul Sabatier TOULOUSE cedex 04 FRANCE

Pricing and hedging in stochastic volatility regime switching models.

Local risk minimizing strategy in a market driven by time-changed Lévy noises. Lotti Meijer Master s Thesis, Autumn 2016

Portfolio Optimization with Nondominated Priors and Unbounded Parameters

Consumption investment optimization with Epstein-Zin utility

Simulation of BSDEs and. Wiener Chaos Expansions

Singular perturbation control problems: a BSDE approach

Examples of Dynamic Programming Problems

Essential Microeconomics : OPTIMAL CONTROL 1. Consider the following class of optimization problems

The Optimal Stopping Time for Selling an Asset When It Is Uncertain Whether the Price Process Is Increasing or Decreasing When the Horizon Is Infinite

Vehicle Arrival Models : Headway

Variational Iteration Method for Solving System of Fractional Order Ordinary Differential Equations

arxiv: v2 [q-fin.mf] 7 Nov 2016

Satisfying Convex Risk Limits by Trading

arxiv: v2 [q-fin.mf] 19 Mar 2017

1 Solutions to selected problems

Lecture Notes 5: Investment

arxiv: v1 [math.pr] 6 Oct 2008

Online Appendix to Solution Methods for Models with Rare Disasters

Dual Representation as Stochastic Differential Games of Backward Stochastic Differential Equations and Dynamic Evaluations

Optimal Investment and Consumption Decisions Under the Ho-Lee Interest Rate Model

Simulation of BSDEs and. Wiener Chaos Expansions

Forward-backward systems for expected utility maximization

Supplement for Stochastic Convex Optimization: Faster Local Growth Implies Faster Global Convergence

Example on p. 157

Nature Neuroscience: doi: /nn Supplementary Figure 1. Spike-count autocorrelations in time.

Loss of martingality in asset price models with lognormal stochastic volatility

Excel-Based Solution Method For The Optimal Policy Of The Hadley And Whittin s Exact Model With Arma Demand

(MS, ) Problem 1

Hamilton Jacobi equations

Institute for Mathematical Methods in Economics. University of Technology Vienna. Singapore, May Manfred Deistler

Lecture 20: Riccati Equations and Least Squares Feedback Control

A note on high-order short-time expansions for ATM option prices under the CGMY model

dy dx = xey (a) y(0) = 2 (b) y(1) = 2.5 SOLUTION: See next page

Introduction to Probability and Statistics Slides 4 Chapter 4

Optimization problem under change of regime of interest rate

On the Timing Option in a Futures Contract

Nonlinear expectations and nonlinear pricing

EMS SCM joint meeting. On stochastic partial differential equations of parabolic type

Investment performance measurement under asymptotically linear local risk tolerance

Stochastic control under progressive enlargement of filtrations and applications to multiple defaults risk management

Exponential utility indifference valuation in two Brownian settings with stochastic correlation

Finance Research Letters. Maximizing utility of consumption subject to a constraint on the probability of lifetime ruin

The Structure of General Mean-Variance Hedging Strategies

PROPERTIES OF MAXIMUM LIKELIHOOD ESTIMATES IN DIFFUSION AND FRACTIONAL-BROWNIAN MODELS

Stochastic Modelling of Electricity and Related Markets: Chapter 3

THE BERNOULLI NUMBERS. t k. = lim. = lim = 1, d t B 1 = lim. 1+e t te t = lim t 0 (e t 1) 2. = lim = 1 2.

Weyl sequences: Asymptotic distributions of the partition lengths

Financial Econometrics Jeffrey R. Russell Midterm Winter 2009 SOLUTIONS

Mean-Variance Hedging for General Claims

An introduction to the theory of SDDP algorithm

Homogenization of random Hamilton Jacobi Bellman Equations

arxiv: v1 [math.pr] 21 May 2010

Modern Dynamic Asset Pricing Models

Lecture 2 April 04, 2018

Time discretization of quadratic and superquadratic Markovian BSDEs with unbounded terminal conditions

THE MYSTERY OF STOCHASTIC MECHANICS. Edward Nelson Department of Mathematics Princeton University

Lecture Notes 2. The Hilbert Space Approach to Time Series

Unit Root Time Series. Univariate random walk

Transcription:

Opimal Porfolio under Fracional Sochasic Environmen Ruimeng Hu Join work wih Jean-Pierre Fouque Deparmen of Saisics and Applied Probabiliy Universiy of California, Sana Barbara Mahemaical Finance Colloquium Universiy of Souhern California Jan 29, 2018 Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 1 / 27

Porfolio Opimizaion: Meron s Problem An invesor manages her porfolio by invesing on a riskless asse B and one risky asse S (single asse for simpliciy) { db = rb d ds = µs d + σs dw π amoun of wealh invesed in he risky asse a ime he wealh process associaed o he sraegy π X π dx π = π ds + Xπ π db (self-financing) S B = (rx π + π (µ r)) d + π σ dw Objecive: M(, x; λ) := sup π A(x,) E [U(X π T ) X π = x where A(x) conains all admissible π and U(x) is a uiliy funcion on R + Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 2 / 27

Porfolio Opimizaion: Meron s Problem An invesor manages her porfolio by invesing on a riskless asse B and one risky asse S (single asse for simpliciy) { db = rb d ds = µs d + σs dw π amoun of wealh invesed in he risky asse a ime he wealh process associaed o he sraegy π X π dx π = π ds + Xπ π db (self-financing) S B = (rx π + π (µ r)) d + π σ dw Objecive: M(, x; λ) := sup π A(x,) E [U(X π T ) X π = x where A(x) conains all admissible π and U(x) is a uiliy funcion on R + Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 2 / 27

Sochasic Volailiy In Meron s work, µ and σ are consan, complee marke Empirical sudies reveal ha σ exhibis random variaion Implied volailiy skew or smile Sochasic volailiy model: µ(y ), σ(y ) incomplee marke Rough Fracional Sochasic volailiy: Gaheral, Jaisson and Rosenbaum 14 Jaisson, Rosenbaum 16 Omar, Masaaki and Rosenbaum 16 We sudy he Meron problem under slowly varying / fas mean-revering fracional sochasic environmen: Nonlinear + Non-Markovian HJB PDE no available Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 3 / 27

Sochasic Volailiy In Meron s work, µ and σ are consan, complee marke Empirical sudies reveal ha σ exhibis random variaion Implied volailiy skew or smile Sochasic volailiy model: µ(y ), σ(y ) incomplee marke Rough Fracional Sochasic volailiy: Gaheral, Jaisson and Rosenbaum 14 Jaisson, Rosenbaum 16 Omar, Masaaki and Rosenbaum 16 We sudy he Meron problem under slowly varying / fas mean-revering fracional sochasic environmen: Nonlinear + Non-Markovian HJB PDE no available Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 3 / 27

Relaed Lieraure Opion Pricing + Markovian modeling: Fouque, Papanicolaou, Sircar and Solna 11 (CUP) Porfolio Opimizaion + Markovian modeling: Fouque, Sircar and Zariphopoulou 13 (MF) Fouque and H. 16 (SICON) Opion Pricing + Non-Markovian modeling: Garnier and Solna 15 (SIFIN), 16 (MF) Porfolio Opimizaion + Non-Markovian modeling: Fouque and H. (slow facor, under revision a MF) Fouque and H. (fas facor, under revision a SIFIN) Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 4 / 27

A General Non-Markovian Model Dynamics of he risky asse S { ds = S [µ(y ) d + σ(y ) dw, { (W Y : a general sochasic process, G := σ Y ) } -adaped, 0 u wih d W, W Y = ρ d. Dynamics of he wealh process X (assume r = 0 for simpliciy): dx π = π µ(y ) d + π σ(y ) dw Define he value process V by V := ess sup π A E [U(X π T ) F where U(x) is of power ype U(x) = x1 γ 1 γ, γ > 0. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 5 / 27

If Y is Markovian... For example, Y is a diffusion process dy = k(y ) d + h(y ) dw Y, V V (, x, y) characerized by a nonlinear HJB PDE Wih a disorion ransformaion 1 V (, x, y) = x1 γ 1 γ Ψ(, y)q, Ψ solves he linear PDE ( 1 Ψ + 2 h2 (y) yy + k(y) y + ρ 1 γ γ λ(y)h(y) y and has he probabilisic represenaion Ψ(, y) = [e Ẽ 1 γ T 2qγ λ 2 (Y s) ds Y = y. 1 Zariphopoulou 99 ) Ψ + 1 γ 2qγ λ2 (y)ψ = 0, Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 6 / 27

If Y is Markovian... For example, Y is a diffusion process dy = k(y ) d + h(y ) dw Y, V V (, x, y) characerized by a nonlinear HJB PDE Wih a disorion ransformaion 1 V (, x, y) = x1 γ 1 γ Ψ(, y)q, Ψ solves he linear PDE ( 1 Ψ + 2 h2 (y) yy + k(y) y + ρ 1 γ γ λ(y)h(y) y and has he probabilisic represenaion Ψ(, y) = [e Ẽ 1 γ T 2qγ λ 2 (Y s) ds Y = y. 1 Zariphopoulou 99 ) Ψ + 1 γ 2qγ λ2 (y)ψ = 0, Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 6 / 27

If Y is Markovian... For example, Y is a diffusion process dy = k(y ) d + h(y ) dw Y, V V (, x, y) characerized by a nonlinear HJB PDE Wih a disorion ransformaion 1 V (, x, y) = x1 γ 1 γ Ψ(, y)q, Ψ solves he linear PDE ( 1 Ψ + 2 h2 (y) yy + k(y) y + ρ 1 γ γ λ(y)h(y) y and has he probabilisic represenaion Ψ(, y) = [e Ẽ 1 γ T 2qγ λ 2 (Y s) ds Y = y. 1 Zariphopoulou 99 ) Ψ + 1 γ 2qγ λ2 (y)ψ = 0, Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 6 / 27

If Y is Markovian + Slowly Varying... For example, Y is a diffusion process dy = δk(y ) d + δh(y ) dw Y, V V (, x, y) characerized by a nonlinear HJB PDE Wih a disorion ransformaion 1 V (, x, y) = x1 γ 1 γ Ψ(, y)q, Ψ solves he linear PDE ( 1 Ψ + 2 δh2 (y) yy + δk(y) y + δρ 1 γ γ λ(y)h(y) y and has he expansion Ψ(, y) = ψ (0) (, y) + δψ (1) (, y) + δψ (2) (, y) +. 1 Zariphopoulou 99 ) Ψ+ 1 γ 2qγ λ2 (y)ψ = 0 Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 6 / 27

In General: Maringale Disorion Transformaion 2 The value process V is given by V = X1 γ [Ẽ (e 1 γ T ) 2qγ λ 2 (Y s) ds q µ(y) G, λ(y) = 1 γ σ(y) where under P, W Y := W Y + 0 a s ds is a sandard BM. The opimal sraegy π is [ π λ(y ) = γσ(y ) + ρqξ X γσ(y ) where ξ is given by he maringale represenaion dm = M ξ d W Y and M is M = [e Ẽ 1 γ T 2qγ 0 λ2 (Y s) ds G 2 Tehranchi 04: differen uiliy funcion, proof and assumpions Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 7 / 27

Remarks only works for one facor model assumpions: inegrabiliy condiions of ξ, X π and π γ = 1 case of log uiliy, can be reaed separaely degenerae case λ(y) = λ 0, M is a consan maringale, ξ = 0 V = X1 γ 1 γ e 1 γ 2γ λ2 0 (T ), π = λ 0 γσ(y ) X. uncorrelaed case ρ = 0, he problem is linear since q = 1 V = X1 γ [ 1 γ E e 1 γ T 2γ λ 2 (Y s) ds G, π = λ(y ) γσ(y ) X. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 8 / 27

Skech of Proof (Verificaion) V is a supermaringale for any admissible conrol π V is a rue maringale following π π is admissible Define α = π /X, hen wih he drif facor D (α ) dv = V D (α ) d + d Maringale D (α ) := α µ γ 2 α2 σ 2 λ2 2γ + q 1 γ a q(q 1) ξ + 2(1 γ) ξ2 + ρqα σξ. α and D (α ) = 0 wih he righ choice of a and q: ( ) 1 γ γ a = ρ λ(y ), q = γ γ + (1 γ)ρ 2. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 9 / 27

Skech of Proof (Verificaion) V is a supermaringale for any admissible conrol π V is a rue maringale following π π is admissible Define α = π /X, hen wih he drif facor D (α ) dv = V D (α ) d + d Maringale D (α ) := α µ γ 2 α2 σ 2 λ2 2γ + q 1 γ a q(q 1) ξ + 2(1 γ) ξ2 + ρqα σξ. α and D (α ) = 0 wih he righ choice of a and q: ( ) 1 γ γ a = ρ λ(y ), q = γ γ + (1 γ)ρ 2. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 9 / 27

Skech of Proof (Verificaion) V is a supermaringale for any admissible conrol π V is a rue maringale following π π is admissible Define α = π /X, hen wih he drif facor D (α ) dv = V D (α ) d + d Maringale D (α ) := α µ γ 2 α2 σ 2 λ2 2γ + q 1 γ a q(q 1) ξ + 2(1 γ) ξ2 + ρqα σξ. α and D (α ) = 0 wih he righ choice of a and q: ( ) 1 γ γ a = ρ λ(y ), q = γ γ + (1 γ)ρ 2. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 9 / 27

Muliple Asses Modeling Consider he following model of S 1, S 2,..., S n ds i = µ i (Y i )S i d + n j=1 σ ij (Y i )S i dw j, i = 1, 2,... n. Each S i is driven by a sochasic facor Y i, bu all facors Y i are adaped o he same single Brownian moion W Y wih he correlaion srucure: d W i, W j = 0, d W i, W Y = ρ d, i, j = 1, 2,..., n, nρ2 < 1. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 10 / 27

Maringale Disorion Transformaion wih Muliple Asses Then, he porfolio value V can be expressed as V = X1 γ [Ẽ (e 1 γ T ) 2qγ µ(y s) Σ(Y s) 1 µ(y s) ds q G, 1 γ he consan q is chosen o be: q = The opimal conrol π is given by [ π Σ(Y ) 1 µ(y ) = γ γ γ + (1 γ)ρ 2 n. + ρqξ σ 1 (Y ) 1 n X. γ Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 11 / 27

Fracional Processes A fracional Brownian moion W (H), H (0, 1) a coninuous Gaussian process zero [ mean E W (H) W s (H) = σ2 H 2 ( 2H + s 2H s 2H) H < 1/2: shor-range correlaion; H > 1/2: long-range correlaion A fracional Ornsein Uhlenbeck process solves dz H = az H d + dw (H) saionary soluion Z H = e a( s) dw s (H) = K( s) dw s Z Gaussian process wih zero mean and consan variance K is non-negaive, K L 2 Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 12 / 27

Fracional Processes A fracional Brownian moion W (H), H (0, 1) a coninuous Gaussian process zero [ mean E W (H) W s (H) = σ2 H 2 ( 2H + s 2H s 2H) H < 1/2: shor-range correlaion; H > 1/2: long-range correlaion A fracional Ornsein Uhlenbeck process solves dz H = az H d + dw (H) saionary soluion Z H = e a( s) dw s (H) = K( s) dw s Z Gaussian process wih zero mean and consan variance K is non-negaive, K L 2 Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 12 / 27

Fracional Processes A fracional Brownian moion W (H), H (0, 1) a coninuous Gaussian process zero [ mean E W (H) W s (H) = σ2 H 2 ( 2H + s 2H s 2H) H < 1/2: shor-range correlaion; H > 1/2: long-range correlaion A fracional Ornsein Uhlenbeck process solves dz H = az H d + dw (H) saionary soluion Z H = e a( s) dw s (H) = K( s) dw s Z Gaussian process wih zero mean and consan variance K is non-negaive, K L 2 Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 12 / 27

Meron Problem under Slowly Varying Fracional SV Consider a rescaled saionary fou process Z δ,h { [ ds = S µ(z δ,h ) d + σ(z δ,h ) dw, Z δ,h = Kδ ( s) dw Z s, K δ () = δk(δ), d W, W Z = ρ d. Our sudy gives, for all H (0, 1): The value process V δ := ess sup π A δ E [U(X π T ) F The corresponding opimal sraegy π Firs order approximaions o V δ and π A pracical sraegy o generae his approximaed value process Apply he maringale disorion ransformaion wih Y = Z δ,h V δ = X1 γ [Ẽ (e 1 γ T ) 2qγ λ 2 (Zs δ,h ) ds q G, 1 γ Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 13 / 27

Meron Problem under Slowly Varying Fracional SV Consider a rescaled saionary fou process Z δ,h { [ ds = S µ(z δ,h ) d + σ(z δ,h ) dw, Z δ,h = Kδ ( s) dw Z s, K δ () = δk(δ), d W, W Z = ρ d. Our sudy gives, for all H (0, 1): The value process V δ := ess sup π A δ E [U(X π T ) F The corresponding opimal sraegy π Firs order approximaions o V δ and π A pracical sraegy o generae his approximaed value process Apply he maringale disorion ransformaion wih Y = Z δ,h V δ = X1 γ [Ẽ (e 1 γ T ) 2qγ λ 2 (Zs δ,h ) ds q G, 1 γ Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 13 / 27

Meron Problem under Slowly Varying Fracional SV Consider a rescaled saionary fou process Z δ,h { [ ds = S µ(z δ,h ) d + σ(z δ,h ) dw, Z δ,h = Kδ ( s) dw Z s, K δ () = δk(δ), d W, W Z = ρ d. Our sudy gives, for all H (0, 1): The value process V δ := ess sup π A δ E [U(X π T ) F The corresponding opimal sraegy π Firs order approximaions o V δ and π A pracical sraegy o generae his approximaed value process Apply he maringale disorion ransformaion wih Y = Z δ,h V δ = X1 γ [Ẽ (e 1 γ T ) 2qγ λ 2 (Zs δ,h ) ds q G, 1 γ Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 13 / 27

Approximaion o he Value Process Theorem (Fouque-H. 17) For fixed [0, T ), he value process V δ akes he form V δ = X1 γ 1 γ e 1 γ 2γ λ2 (Z δ,h 0 )(T ) + X1 γ + δ H ρ X1 γ 1 γ e 1 γ 2γ λ2 (Z δ,h 0 )(T ) λ 2 (Z δ,h 0 )λ (Z δ,h 0 ) + O(δ 2H ), where φ δ is he random componen of order δ H γ [ T ( φ δ = E Zs δ,h e 1 γ 2γ λ2 (Z δ,h 0 )(T ) λ(z δ,h 0 )λ (Z δ,h 0 )φ δ ( ) 1 γ 2 (T ) H+ 3 2 ) Z δ,h 0 γ ds G. Γ(H + 5 2 ) Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 14 / 27

Approximaion o he Opimal Sraegy Recall ha [ π λ(z δ,h ) = γσ(z δ,h ) + ρqξ γσ(z δ,h X ) and ξ is from he maringale rep. of M = [e Ẽ 1 γ T 2qγ 0 λ2 (Zs δ,h ) ds G. Theorem (Fouque-H., 17) The opimal sraegy π is approximaed by [ π λ(z δ,h ) ρ(1 γ) (T ) H+1/2 = γσ(z δ,h + δh ) γ 2 σ(z δ,h ) Γ(H + 3 2 ) λ(z δ,h 0 )λ (Z δ,h 0 ) + O(δ 2H ) := π (0) + δ H π (1) + O(δ 2H ). X Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 15 / 27

Approximaion o he Opimal Sraegy Recall ha [ π λ(z δ,h ) = γσ(z δ,h ) + ρqξ γσ(z δ,h X ) and ξ is from he maringale rep. of M = [e Ẽ 1 γ T 2qγ 0 λ2 (Zs δ,h ) ds G. Theorem (Fouque-H., 17) The opimal sraegy π is approximaed by [ π λ(z δ,h ) ρ(1 γ) (T ) H+1/2 = γσ(z δ,h + δh ) γ 2 σ(z δ,h ) Γ(H + 3 2 ) λ(z δ,h 0 )λ (Z δ,h 0 ) + O(δ 2H ) := π (0) + δ H π (1) + O(δ 2H ). X Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 15 / 27

How Good is he Approximaion? Corollary In he case of power uiliy U(x) = x1 γ 1 γ, π(0) = λ(zδ,h X γσ(z δ,h generaes he ) approximaion of V δ up o order δ H (leading order + wo correcion erms of order δ H ), hus asympoically opimal in A δ. ) H = 1 2, Zδ,H becomes he Markovian OU process, boh approximaion coincides wih resuls in [Fouque Sircar Zariphopoulou 13. The corollary recovers [Fouque -H. 16. Skech of proofs: Apply Taylor ( expansion ) o λ(z) a he poin Z δ,h 0, and hen conrol he momens Z δ,h Z δ,h 0 δ H. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 16 / 27

How Good is he Approximaion? Corollary In he case of power uiliy U(x) = x1 γ 1 γ, π(0) = λ(zδ,h X γσ(z δ,h generaes he ) approximaion of V δ up o order δ H (leading order + wo correcion erms of order δ H ), hus asympoically opimal in A δ. ) H = 1 2, Zδ,H becomes he Markovian OU process, boh approximaion coincides wih resuls in [Fouque Sircar Zariphopoulou 13. The corollary recovers [Fouque -H. 16. Skech of proofs: Apply Taylor ( expansion ) o λ(z) a he poin Z δ,h 0, and hen conrol he momens Z δ,h Z δ,h 0 δ H. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 16 / 27

Meron Problem under Fas-Varying Fracional SV Consider a ɛ-scaled saionary fou process Y ɛ,h Y ɛ,h = ɛ H e a( s) ɛ dw s (H) = K ɛ ( s) dws Y, K ɛ () = 1 K( ɛ ɛ ) ogeher wih he risky asse [ ds = S µ(y ɛ,h ) d + σ(y ɛ,h ) dw, d W, W Y = ρ d. For power uiliies, we obain: The value process V ɛ and he corresponding opimal sraegy π Firs order approximaions o V ɛ and π A sraegy π (0) o generae his approximaed value process Using ergodic propery of Y ɛ,h, bu only valid for H ( 1 2, 1) Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 17 / 27

Meron Problem under Fas-Varying Fracional SV Consider a ɛ-scaled saionary fou process Y ɛ,h Y ɛ,h = ɛ H e a( s) ɛ dw s (H) = K ɛ ( s) dws Y, K ɛ () = 1 K( ɛ ɛ ) ogeher wih he risky asse [ ds = S µ(y ɛ,h ) d + σ(y ɛ,h ) dw, d W, W Y = ρ d. For power uiliies, we obain: The value process V ɛ and he corresponding opimal sraegy π Firs order approximaions o V ɛ and π A sraegy π (0) o generae his approximaed value process Using ergodic propery of Y ɛ,h, bu only valid for H ( 1 2, 1) Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 17 / 27

Meron Problem under Fas-Varying Fracional SV Consider a ɛ-scaled saionary fou process Y ɛ,h Y ɛ,h = ɛ H e a( s) ɛ dw s (H) = K ɛ ( s) dws Y, K ɛ () = 1 K( ɛ ɛ ) ogeher wih he risky asse [ ds = S µ(y ɛ,h ) d + σ(y ɛ,h ) dw, d W, W Y = ρ d. For power uiliies, we obain: The value process V ɛ and he corresponding opimal sraegy π Firs order approximaions o V ɛ and π A sraegy π (0) o generae his approximaed value process Using ergodic propery of Y ɛ,h, bu only valid for H ( 1 2, 1) Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 17 / 27

Approximaion o he Value Process V ɛ Theorem (Fouque-H. 17) For fixed [0, T ), he value process V ɛ akes he form V ɛ = X1 γ 1 γ e 1 γ 2γ λ2 (T ) + X1 γ + ɛ 1 H ρ X1 γ 1 γ e 1 γ 2γ λ2 (T ) λ + o(ɛ 1 H ), where φ ɛ is he random componen of order ɛ 1 H e 1 γ 2γ λ2 (T ) φ ɛ γ ( ) 1 γ 2 λλ (T ) H+ 1 2 γ aγ(h + 3 2 ) [ 1 T ( φ ɛ = E λ 2 (Ys ɛ,h ) λ 2) ds 2 G. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 18 / 27

Opimal Porfolio Theorem (Fouque-H., 17) The opimal sraegy π is approximaed by [ π λ(y ɛ,h ) ρ(1 γ) λλ (T ) H 1/2 = γσ(y ɛ,h + ɛ1 H ) γ 2 σ(y ɛ,h ) aγ(h + 1 2 ) Corollary + o(ɛ 1 H ) ɛ,h λ(y ) γσ(y ɛ,h In he case of power uiliy, π (0) = X generaes he ) approximaion of V ɛ up o order ɛ 1 H (leading order + wo correcion erms of order ɛ 1 H ), hus asympoically opimal in A ɛ. X Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 19 / 27

Ergodiciy of Y ɛ,h For H > 1/2, under appropriae assumpion of λ( ), 0 0 0 (λ 2 (Y ɛ,h s (λ(y ɛ,h s (λ(y ɛ,h s are small and of order ɛ 1 H. ) λ 2 ) ds, ) λ) ds, )λ (Ys ɛ,h ) λλ ) ds, Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 20 / 27

Comparison wih he Markovian Case The value funcion and he opimal sraegy are derived in [FSZ, 13: [ V ɛ (, X ) = X1 γ 1 γ e 1 γ 2γ λ2 (T ) 1 ( ) 1 γ 2 λθ ɛρ (T ) + O(ɛ) γ 2 [ π (, X, Y ɛ,h λ(y ɛ,h ) ) = γσ(y ɛ,h ) + ρ(1 γ) θ (Y ɛ,h ) ɛ γ 2 σ(y ɛ,h X + O(ɛ) ) 2 Formally le H 1 2 in our resuls: [ V ɛ = X1 γ 1 γ e 1 γ 2γ λ2 (T ) 1+ ( ) 1 γ 2 λ λλ ɛρ (T ) + o( ɛ) γ a [ π λ(y ɛ,h ) = γσ(y ɛ,h ) + ρ(1 γ) λλ ɛ γ 2 σ(y ɛ,h X + o( ɛ) ) a Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 21 / 27

Work in preparaion Rough Fas-varying fsv Surprisingly, Y ɛ,h order correcion: V ɛ Y ɛ,h = = X1 γ 1 γ e 1 γ 2γ λ2 (T ) Muliscale fsv ds = S [ Y ɛ,h = µ(y ɛ,h K ɛ ( s) dw Y s, H < 1 2 is no visible o he leading order nor in he firs, Z δ,h K ɛ ( s) dw Y, [ 1 + ( ) 1 γ 2 ɛρ D(T ) + o( ɛ) γ ) d + σ(y ɛ,h Z δ,h =, Z δ,h ) dw, K δ ( s) dw Z. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 22 / 27

Work in preparaion Rough Fas-varying fsv Surprisingly, Y ɛ,h order correcion: V ɛ Y ɛ,h = = X1 γ 1 γ e 1 γ 2γ λ2 (T ) Muliscale fsv ds = S [ Y ɛ,h = µ(y ɛ,h K ɛ ( s) dw Y s, H < 1 2 is no visible o he leading order nor in he firs, Z δ,h K ɛ ( s) dw Y, [ 1 + ( ) 1 γ 2 ɛρ D(T ) + o( ɛ) γ ) d + σ(y ɛ,h Z δ,h =, Z δ,h ) dw, K δ ( s) dw Z. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 22 / 27

Meron Problem under General Uiliy Maringale Disorion Transformaion is no available Sar wih a given sraegy π (0) A firs order approximaion o V π(0),δ obained by epsilon-maringale decomposiion 34 Opimaliy of π (0) in a smaller class of conrols of feedback form Denoe by v (0) (, x, z) he value funcion a he Sharpe-raio λ(z), we define π (0) by π (0) (, x, z) = λ(z) v x (0) (, x, z) σ(z) v xx (0) (, x, z) and he associaed value process V π(0),δ V π(0),δ 3 Fouque Papanicolaou Sircar 01 4 Garnier Solna 15 := E [ U(X π(0) T ) F. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 23 / 27

Meron Problem under General Uiliy Maringale Disorion Transformaion is no available Sar wih a given sraegy π (0) A firs order approximaion o V π(0),δ obained by epsilon-maringale decomposiion 34 Opimaliy of π (0) in a smaller class of conrols of feedback form Denoe by v (0) (, x, z) he value funcion a he Sharpe-raio λ(z), we define π (0) by π (0) (, x, z) = λ(z) v x (0) (, x, z) σ(z) v xx (0) (, x, z) and he associaed value process V π(0),δ V π(0),δ 3 Fouque Papanicolaou Sircar 01 4 Garnier Solna 15 := E [ U(X π(0) T ) F. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 23 / 27

Meron Problem under General Uiliy Maringale Disorion Transformaion is no available Sar wih a given sraegy π (0) A firs order approximaion o V π(0),δ obained by epsilon-maringale decomposiion 34 Opimaliy of π (0) in a smaller class of conrols of feedback form Denoe by v (0) (, x, z) he value funcion a he Sharpe-raio λ(z), we define π (0) by π (0) (, x, z) = λ(z) v x (0) (, x, z) σ(z) v xx (0) (, x, z) and he associaed value process V π(0),δ V π(0),δ 3 Fouque Papanicolaou Sircar 01 4 Garnier Solna 15 := E [ U(X π(0) T ) F. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 23 / 27

Meron Problem under General Uiliy Maringale Disorion Transformaion is no available Sar wih a given sraegy π (0) A firs order approximaion o V π(0),δ obained by epsilon-maringale decomposiion 34 Opimaliy of π (0) in a smaller class of conrols of feedback form Denoe by v (0) (, x, z) he value funcion a he Sharpe-raio λ(z), we define π (0) by π (0) (, x, z) = λ(z) v x (0) (, x, z) σ(z) v xx (0) (, x, z) and he associaed value process V π(0),δ V π(0),δ 3 Fouque Papanicolaou Sircar 01 4 Garnier Solna 15 := E [ U(X π(0) T ) F. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 23 / 27

Epsilon-Maringale Decomposiion Finding Q π(0),δ Then such ha Q π(0),δ T = V π(0),δ T = U(XT π(0) ), Q π(0),δ = M δ + R δ, where M δ is a maringale and R δ is of order δ 2H. V π(0),δ [ = E Q π(0),δ T F = Q π(0),δ R δ + E = M δ + E [RT δ F [R δ T F, and Q π(0),δ is he approximaion o V π(0),δ wih error of order O(δ 2H ) Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 24 / 27

Epsilon-Maringale Decomposiion Finding Q π(0),δ Then such ha Q π(0),δ T = V π(0),δ T = U(XT π(0) ), Q π(0),δ = M δ + R δ, where M δ is a maringale and R δ is of order δ 2H. V π(0),δ [ = E Q π(0),δ T F = Q π(0),δ R δ + E = M δ + E [RT δ F [R δ T F, and Q π(0),δ is he approximaion o V π(0),δ wih error of order O(δ 2H ) Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 24 / 27

Epsilon-Maringale Decomposiion Finding Q π(0),δ Then such ha Q π(0),δ T = V π(0),δ T = U(XT π(0) ), Q π(0),δ = M δ + R δ, where M δ is a maringale and R δ is of order δ 2H. V π(0),δ [ = E Q π(0),δ T F = Q π(0),δ R δ + E = M δ + E [RT δ F [R δ T F, and Q π(0),δ is he approximaion o V π(0),δ wih error of order O(δ 2H ) Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 24 / 27

Firs order approximaion o V π(0),δ Proposiion For fixed [0, T ), he F -measurable value process V π(0),δ is of he form V π(0),δ = Q π(0),δ where Q π(0),δ (x, z) is given by: (X π(0), Z δ,h 0 ) + O(δ 2H ), Q π(0),δ (x, z) =v (0) (, x, z) + λ(z)λ (z)d 1 v (0) (, x, z)φ δ + δ H ρλ 2 (z)λ (z)d1v 2 (0) (T )H+3/2 (, x, z) Γ(H + 5 2 ). For power uiliy, Q π(0),δ coincides wih he approximaion of V δ For he Markovian case H = 1 2, recovers he resuls in [Fouque-H. 16 Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 25 / 27

Firs order approximaion o V π(0),δ Proposiion For fixed [0, T ), he F -measurable value process V π(0),δ is of he form V π(0),δ = Q π(0),δ where Q π(0),δ (x, z) is given by: (X π(0), Z δ,h 0 ) + O(δ 2H ), Q π(0),δ (x, z) =v (0) (, x, z) + λ(z)λ (z)d 1 v (0) (, x, z)φ δ + δ H ρλ 2 (z)λ (z)d1v 2 (0) (T )H+3/2 (, x, z) Γ(H + 5 2 ). For power uiliy, Q π(0),δ coincides wih he approximaion of V δ For he Markovian case H = 1 2, recovers he resuls in [Fouque-H. 16 Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 25 / 27

Asympoically Opimaliy of π (0) Theorem (Fouque-H. 17) The rading sraegy π (0) (, x, z) = λ(z) σ(z) v x (0) (,x,z) v xx (0) (,x,z) is asympoically opimal in he following class: { Ã δ [ π 0, π 1, α := π = π 0 + δ α π } 1 : π A δ, α > 0, 0 < δ 1. Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 26 / 27

Thank you! Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 27 / 27

Sochasic Volailiy wih Fas Facor Y S is modeled by: { ds = µ(y )S d + σ(y )S dw, dy = 1 ɛ b(y ) d + 1 ɛ a(y ) dw Y, wih correlaion dw W Y = ρ d. Theorem (Fouque-H., in prep.) Under appropriae assumpions, for fixed (, x, y) and any family of rading sraegies A 0 (, x, y) [ π 0, π 1, α, he following limi exiss and saisfies Ṽ ɛ (, x, y) V π(0),ɛ (, x, y) l := lim 0. ɛ 0 ɛ Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 27 / 27

Theorem (Fouque-H., in prep.) The residual funcion E(, x, y) := V π(0),ɛ (, x) v (0) (, x) ɛv (1) (, x) is of order ɛ, where in his case, v (0) solves v (0) 1 2 λ2 and v (1) = 1 2 (T )ρ 1BD 2 1 v(0) (, x). ( v (0) x v (0) xx ) 2 = 0, Ruimeng Hu (UCSB) Opimal Porfolio under Fracional Environmen Jan. 29, 2018 27 / 27