KennethL.Judd,HooverInstitution YongyangCai...

Dynamic Programming for Portfolio Problems

Kenneth L. Judd, Hoover InstitutionYongyang Cai, Hoover Institution

May 27, 2012

Proportional Transaction Cost and CRRA Utility

I Separability of wealth W and portfolio fractions x .I If u(W ) = W 1−γ/(1− γ), then Vt(Wt , xt) = W 1−γ

t · gt(xt).I If u(W ) = log(W ), then Vt(Wt , xt) = log(Wt) + ψt(xt).I “No-trade” region: Ωt = xt : (δ+t )∗ = (δ−t )∗ = 0, where

(δ+t )∗ ≥ 0 are fractions of wealth for buying stocks, and (δ−t )∗ ≥ 0are fractions of wealth for selling stocks.

Examples with two stocks

We first look at simple examples:I Two stocks and one bondI Investment begins at t = 0 and is liquidated at t = 6

i.i.d. returns (uniform distribution on [0.87,1.27])

0.5 0.55 0.6 0.65 0.7 0.75 0.80.5

0.55

0.6

0.65

0.7

0.75

0.8

Stock 1 fraction

Sto

ck 2

fra

ctio

nNo−trade region at stage t=0

0.5 0.55 0.6 0.65 0.7 0.75 0.80.5

0.55

0.6

0.65

0.7

0.75

0.8

Stock 1 fraction

Sto

ck 2

fra

ctio

n

No−trade region at stage t=3

0.5 0.55 0.6 0.65 0.7 0.75 0.80.5

0.55

0.6

0.65

0.7

0.75

0.8

Stock 1 fraction

Sto

ck 2

fra

ctio

n

No−trade region at stage t=5

Correlated returns (discrete)

I Return: R1 = 0.85, 1.08, 1.25; R2 = 0.88, 1.06, 1.2;

I Joint Probability Matrix:

0.08 0.10.12 0.4 0.12

0.1 0.08

0.3 0.4 0.5 0.6 0.7 0.8 0.9−0.5

−0.4

−0.3

−0.2

−0.1

0

0.1

Stock 1 fraction

Sto

ck 2

fra

ctio

n

No−trade region at stage t=0

0.3 0.4 0.5 0.6 0.7 0.8 0.9−0.5

−0.4

−0.3

−0.2

−0.1

0

0.1

Stock 1 fraction

Sto

ck 2

fra

ctio

n

No−trade region at stage t=3

0.3 0.4 0.5 0.6 0.7 0.8 0.9−0.5

−0.4

−0.3

−0.2

−0.1

0

0.1

Stock 1 fraction

Sto

ck 2

fra

ctio

n

No−trade region at stage t=5

Stochastic mean return

I Log-normal return: N(µi − σ2

i /2, σ2i

)with µi = 0.06 or 0.08 and

σi = 0.2, for i = 1, 2

I Transition probability matrix of µ:[

0.75 0.250.25 0.75

]

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

Stock 1 fraction

Sto

ck 2

fra

ctio

n

No−trade region at stage t=0

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

Stock 1 fraction

Sto

ck 2

fra

ctio

n

No−trade region at stage t=3

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

Stock 1 fraction

Sto

ck 2

fra

ctio

n

No−trade region at stage t=5

Three stocks

I Problem: 3 stocks + 1 bond.I Investment begins at t = 0 and is liquidated at t = 6I Log-normal return: N

(µ− σ2/2, ΛΣΛ

)with σ = (0.16, 0.18, 0.2)>,

µ = (0.07, 0.08, 0.09)>, Λ is the diagonal matrix of σ, and

Σ =

1 0.2 0.10.2 1 0.3140.1 0.314 1

.I degree-7 complete Chebyshev approximationI product Gauss-Hermite quadrature rule with 9 nodes in each

dimension

0.1

0.2

0.3

0.4

0.1

0.2

0.3

0.4

0.1

0.2

0.3

0.4

0.1

0.2

0.3

0.4

0.1

0.2

0.3

0.4

0.1

0.2

0.3

0.4

0.1

0.2

0.3

0.4

0.1

0.2

0.3

0.4

0.1

0.2

0.3

0.4

Portfolio with Transaction Costs and Consumption

I Problem: 3 stocks + 1 bond, 50 periodsI No-trade regions:

Ω49 ≈ [0.16, 0.33]3,Ω48 ≈ [0.14, 0.24]3, . . . ,

Ωt ≈ [0.19, 0.27]3, for t = 0, . . . , 15,

I Merton’s ratio: x∗ = (ΛΣΛ)−1(µ− r)/γ = (0.25, 0.25, 0.25)>

Σ12 = Σ13 = 0.2, Σ23 = 0.04

Σ12 = Σ13 = 0.2, Σ23 = 0.04

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

More correlation: Σ12 = Σ13 = 0.4, Σ23 = 0.16

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

Portfolios with options

I Problem: 1 stock, 1 put option on the stock, and 1 bondI Investment begins at t = 0 and is liquidated at t = 6 monthsI Proportional transaction costs in stock and option tradesI Return of option is based on pricing of the option.

I We use the binomial lattice method to price the optionI Stock price will be one exogenous state variable

I Separability of wealth W and portfolio fractions x and stock price SI If u(W ) = W 1−γ/(1 − γ), then Vt(Wt , St , xt) = W 1−γ

t · gt(St , xt).I If u(W ) = log(W ), then Vt(Wt ,St , xt) = log(Wt) + ψt(St , xt).

1 stock and 1 at-the-money put option at t = 0 (liquidate at t = 6months)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1No−Trade Region, Option Price=0.0462351K

Stock fraction

Option fra

ction

I Put option: strike K , expiration time T , payoff max(K − ST , 0)

I stock price S , utility u(W ) = −W−2/2

Dependence on transaction costs1 at-the-money put option at t = 0

transaction cost ratios: τ1 = 0.01, τ2 = 0.02

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

Stock fraction

Option fra

ction

1 at-the-money put option at t = 0transaction cost ratios: τ1 = 0.01, τ2 = 0.005

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

Stock fraction

Option fra

ction

1 at-the-money put option at t = 0 (liquidate at t = 6)transaction cost ratios: τ1 = 0.005, τ2 = 0.01

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

Stock fraction

Option fra

ction

1 at-the-money put option at t = 0transaction cost ratios: τ1 = 0.005, τ2 = 0.005

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

Stock fraction

Option fra

ction

Dependence on horizon1 at-the-money put option (τ1 = 0.01, τ2 = 0.01)

expiration time: T = 1 month

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Stock fraction

Option fra

ction

1 at-the-money put option (τ1 = 0.01, τ2 = 0.01)expiration time: T = 2 months

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Stock fraction

Option fra

ction

1 at-the-money put option (τ1 = 0.01, τ2 = 0.01)expiration time: T = 3 months

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Stock fraction

Option fra

ction

1 at-the-money put option (τ1 = 0.01, τ2 = 0.01)expiration time: T = 4 months

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Stock fraction

Option fra

ction

1 at-the-money put option (τ1 = 0.01, τ2 = 0.01)expiration time: T = 5 months

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Stock fraction

Option fra

ction

Dependence on strike price (K = 0.8S0)

1 put option at t = 0 (liquidate at t = 6, τ1 = τ2 = 0.01)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0.02No−Trade Region, Option Price=0.00266625K

Stock fraction

Option fra

ction

K = 1.2S0

1 put option at t = 0 (liquidate at t = 6, τ1 = τ2 = 0.01)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

0.18

0.2No−Trade Region, Option Price=0.154645K

Stock fraction

Option fra

ction

Social value of optionsValue functions with/without options at t = 0 (liquidate at t = 6)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1−0.483

−0.482

−0.481

−0.48

−0.479

−0.478

−0.477

−0.476

−0.475

−0.474

Value Functions V0(W,S,x,y) at W=1, S=1 and y=0

Stock fraction x

Valu

e

Option unavailableOption available, τ

2=0.02

Option available, τ2=0.01

Option available, τ2=0.005

I (x , y): fractions of money in stock and optionI τ1 = 0.01 and τ2: transaction cost ratios of stock and option

Summary

I Developed a NDP method with shape-preservingapproximation, stabler

I Developed a NDP method with Hermiteinterpolation, more accurate and more time-saving

I Developed a parallel NDP algorithm, running overhundreds of computers, almost linear speed-up

I Solved arbitrary-number-of-period and many-assetdynamic portfolio problems with transaction costs

I Solved arbitrary-number-of-period dynamicportfolio problems with options and transactioncosts