Скачать презентацию 1 Dynamic portfolio optimization with stochastic programming TIØ Скачать презентацию 1 Dynamic portfolio optimization with stochastic programming TIØ

112981d3323114fba3a7d99ec2396b9c.ppt

  • Количество слайдов: 49

1 Dynamic portfolio optimization with stochastic programming TIØ 4317, H 2009 1 Dynamic portfolio optimization with stochastic programming TIØ 4317, H 2009

2 Dynamic Trading Strategies A sequence of buy and sell decisions, including short term 2 Dynamic Trading Strategies A sequence of buy and sell decisions, including short term borrowing and lending Rebalancing portfolio weights at discrete times • Simple decisions rules for portfolio rebalancing • Stochastic dedication • Stochastic linear programming

3 Modeling portfolio decisions in discrete time • Portfolio decisions can be made at 3 Modeling portfolio decisions in discrete time • Portfolio decisions can be made at a finite number of points in time called trading dates. – No decisions are assumed to be taken between one trading date and the other • Prices over time are modeled following the structure of a binary lattice. • Let us denote – s as the index of a possible state – as the set of the state indices at time t

4 A binary lattice 4 A binary lattice

5 Linear scenario structures • Scenario – A particular scenario is denoted by set 5 Linear scenario structures • Scenario – A particular scenario is denoted by set of the stages in t+1 that can be reached by a state in t set of the stages in t that can reach a particular stage in t+1

6 Linear scenario structure 6 Linear scenario structure

7 Non-anticipativity • Trading strategies can not depend on what happens in the future 7 Non-anticipativity • Trading strategies can not depend on what happens in the future • Two scenarios with the same history up to time require the same strategy to be implemented up to that time (nonanticipativity) • To model non-anticipativity we can recombine the scenarios on an event tree • With a binary lattice every node has two predecessors – We do not know the node we come from • With an event tree we have just an predecessor for each node – We always know the history of the process

8 Event Tree 8 Event Tree

9 Some formal definitions • Event Tree: directed graph • is the set of 9 Some formal definitions • Event Tree: directed graph • is the set of the nodes • is the set of the possible links • is the set of possible states at time t

10 Event tree properties • 1 • 2 Every state has a unique predecessor 10 Event tree properties • 1 • 2 Every state has a unique predecessor

11 Scenarios: formal definition such that for each last trading date for the scenario 11 Scenarios: formal definition such that for each last trading date for the scenario strategy l

12 Decision rules for Dynamic portfolio strategies • • Buy-and-hold Constant mix Constant proportion 12 Decision rules for Dynamic portfolio strategies • • Buy-and-hold Constant mix Constant proportion Option based portfolio insurance – Simple decision rules for rebalancing of portfolios – No optimality – Easy to specify and compute

13 Buy-and-hold • Specify the proportion of the initial wealth invested in the risk 13 Buy-and-hold • Specify the proportion of the initial wealth invested in the risk free and in the risky asset at time 0. – The portfolio is held until maturity under all the scenarios • Let us define the growth of the risky asset value with fixed since time 0 • This portfolio has a minimum value of

14 Constant mix Strategy • Specifies that the proportion of value of the risk 14 Constant mix Strategy • Specifies that the proportion of value of the risk free and the risky asset wrt the portfolio value remains constant for all scenarios/trading times • Values of risky and risk free assets in the portfolio at time t, scenario s

15 Constant mix Strategy (Cont’d…) • Constant mix strategy rebalancing condition 15 Constant mix Strategy (Cont’d…) • Constant mix strategy rebalancing condition

16 Constant proportion strategy • A fixed proportion of the portfolio is invested in 16 Constant proportion strategy • A fixed proportion of the portfolio is invested in the risky asset • This proportion stays fixed all along the lifetime of the investment by rebalancing • This strategy provides for a floor g, below which the asset value is not allowed to fall.

17 Option-based portfolio insurance • A mix of risk free and risky asset such 17 Option-based portfolio insurance • A mix of risk free and risky asset such that the payoff scheme matches the one of a portfolio composed of risk free assets and call options • The risk free assets are kept equal to the floor of the portfolio and any excess value is invested in call options. • If the portfolio value drops down its minimum value is given by the value of the risk free investment.

18 Stochastic dedication • The model optimizes short term borrowing and lending decisions as 18 Stochastic dedication • The model optimizes short term borrowing and lending decisions as new information arrives • It does not account for portfolio rebalancing • Portfolio decisions are optimized at time 0 – Together with the borrowing-lending decisions using the surplusshortage between assets and liabilities.

19 Necessary conditions for immunization: definitions • Discount factor • Present value of asset 19 Necessary conditions for immunization: definitions • Discount factor • Present value of asset i in scenario l • Present value of liabilities in scenario l

20 Necessary conditions for immunization • Necessary condition for scenario immunization • This condition 20 Necessary conditions for immunization • Necessary condition for scenario immunization • This condition can be very expensive or even impossible to satisfy for all scenarios

21 Relaxation • What the model seeks to find is a trade-off between reward, 21 Relaxation • What the model seeks to find is a trade-off between reward, when the asset portfolio outperforms the liabilities against the risk when the portfolio underperforms. • Present value of the asset-liabilities portfolio in scenario l

22 Relaxation (cont’d…) • Define as the initial budget and accepted as the maximum 22 Relaxation (cont’d…) • Define as the initial budget and accepted as the maximum risk

23 Trade-off formulation 23 Trade-off formulation

24 about the immunization • To satisfy the immunization condition we have to include 24 about the immunization • To satisfy the immunization condition we have to include borrowing and lending decisions • The price of a portfolio of assets will be covered by liabilities and loans

25 Cashflow matching • The stochastic cashflow matching equation, encompassing borrowing and lending decisions 25 Cashflow matching • The stochastic cashflow matching equation, encompassing borrowing and lending decisions is given by CF+interests+borrowed funds = liabilities+lended money+ debts

26 Stochastic dedication 26 Stochastic dedication

27 A primer in Stochastic Programming • We are interested in finding a “solution” 27 A primer in Stochastic Programming • We are interested in finding a “solution” to the problem • Idea of the solution: the best value which satisfies the constraints most of the time (a reliable optimal solution)

28 A primer in Stochastic Programming • To find a solution we make suitable 28 A primer in Stochastic Programming • To find a solution we make suitable transformations of the functions in the system – EXAMPLE An optimization problem can be modified with a penalty function for the constraints. The idea with SP is the same: modify the functions in a reasonable way to find a solution

29 A primer in Stochastic Programming • Then solve the problem 29 A primer in Stochastic Programming • Then solve the problem

30 A primer in Stochastic Programming • A typical SP problem 30 A primer in Stochastic Programming • A typical SP problem

31 A primer in Stochastic Programming A possible modification: • Define a variable y 31 A primer in Stochastic Programming A possible modification: • Define a variable y and the function • Now define And drop all the stochastic constraints in the original problem

32 A primer in Stochastic Programming • We obtain the so called recourse formulation 32 A primer in Stochastic Programming • We obtain the so called recourse formulation

33 A primer in Stochastic Programming • If we consider the discrete case we 33 A primer in Stochastic Programming • If we consider the discrete case we can link to a scenario every possible realization of the random variable • In this case we can write the large scale deterministic equivalent

34 A primer in Stochastic Programming • Large scale deterministic equivalent 34 A primer in Stochastic Programming • Large scale deterministic equivalent

35 A primer in Stochastic Programming • If it could be possible to forecast 35 A primer in Stochastic Programming • If it could be possible to forecast the future we could have a different first stage decision for each scenario

36 A primer in Stochastic Programming • Since we can not forecast the future 36 A primer in Stochastic Programming • Since we can not forecast the future we need to enforce nonanticipativity, setting • With nonanticipativity constraints we require that the decisions taken in different scenarios that at a given stage “look the same” have to coincide • This is what is automatically done in the recourse formulation, setting a unique first stage decision

37 A primer in Stochastic Programming • Multistage recourse problems • Recourse on the 37 A primer in Stochastic Programming • Multistage recourse problems • Recourse on the recourse… • It is an extension of the two stage model and it is formulated with a nested structure

38 A primer in Stochastic Programming • Multistage recourse formulation 38 A primer in Stochastic Programming • Multistage recourse formulation

39 A rigorous framework for optimization under uncertainty Stochastic programming • Is the mathematical 39 A rigorous framework for optimization under uncertainty Stochastic programming • Is the mathematical programming tool that facilitates the optimization of dynamic strategies on event trees • Models are optimal and satisfy non-anticipativity • Portfolio rebalancing is allowed as new information becomes available.

40 Stochastic Programming for dynamic strategies • At each trading date the manager assesses 40 Stochastic Programming for dynamic strategies • At each trading date the manager assesses the market conditions • The manager also assesses the potential changes in conditions of market parameters • The new information is incorporated in a sequence of transactions

41 Model formulation • The model encompasses two types of constraints: 1. Inventory balance 41 Model formulation • The model encompasses two types of constraints: 1. Inventory balance constraints 2. Cashflow balance constraints • The model encompasses two levels of constraints: – First stage constraints – Time-staged constraints

42 First stage constraints 1. Inventory balance constraint Face value of assets in the 42 First stage constraints 1. Inventory balance constraint Face value of assets in the portfolio equal to what we had in the portfolio plus what we have bought minus what we have sold 2. Cashflow balance equation Inflows from sales of securities plus borrowed money equal the amount invested for purchasing new securities plus the amount invested in the riskless asset plus the payment of liabilities

43 First stage constraints (Cont’d…) • Inventory balance • Cash flow balance 43 First stage constraints (Cont’d…) • Inventory balance • Cash flow balance

44 Time staged constraints • At each time period we have a set of 44 Time staged constraints • At each time period we have a set of constraints for each scenario • Decisions are conditioned by the state of the system at time t as well as the decisions taken in t-1 at the predecessor state

45 Time staged constraints (Cont’d…) 1. Inventory balance constraints One constraint for each security 45 Time staged constraints (Cont’d…) 1. Inventory balance constraints One constraint for each security and for each scenario 1. Cashflow balance constraints One constraint for each scenario

46 End of horizon constraint • We evaluate the terminal wealth as sum of 46 End of horizon constraint • We evaluate the terminal wealth as sum of the market value of the portfolio of assets and the money the investor has lended.

47 Objective function • The objective function is expressed in form of expected utility 47 Objective function • The objective function is expressed in form of expected utility of terminal wealth

48 Stochastic program for dynamic strategies 48 Stochastic program for dynamic strategies

49 Use of stochastic programming for dynamic portfolio management • The advantage of using 49 Use of stochastic programming for dynamic portfolio management • The advantage of using a stochastic programming framework is that a set of restriction such as transaction costs, multiple state variables, market incompleteness, taxes and trading limits can be handled simultaneously within the framework • The drawback is that the computational effort explodes as the number of scenarios and decision stages increases.