1
FINANCIAL RISKFINANCIAL RISK
CHE 5480CHE 5480Miguel BagajewiczMiguel Bagajewicz
University of OklahomaUniversity of OklahomaSchool of Chemical Engineering and Materials ScienceSchool of Chemical Engineering and Materials Science
2
Scope of DiscussionScope of Discussion
We will discuss the definition and management of financial risk in We will discuss the definition and management of financial risk in in any design process or decision making paradigm, like…in any design process or decision making paradigm, like…
Extensions that are emerging are the treatment of other risksExtensions that are emerging are the treatment of other risks in a multiobjective (?) framework, including for examplein a multiobjective (?) framework, including for example
• Investment PlanningInvestment Planning• Scheduling and more in general, operations planningScheduling and more in general, operations planning• Supply Chain modeling, scheduling and controlSupply Chain modeling, scheduling and control• Short term scheduling (including cash flow management)Short term scheduling (including cash flow management)• Design of process systems Design of process systems • Product DesignProduct Design
• Environmental RisksEnvironmental Risks• Accident Risks (other than those than can be expressed Accident Risks (other than those than can be expressed as financial risk)as financial risk)
3
Introduction – Understanding RiskIntroduction – Understanding Risk
Profit Histogram
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
-300 -200 -100 0 100 200 300 400 500 600 700 800Profit (M$)
Probability
Investment Plan I - E[Profit ] = 338Investment Plan II - E[Profit ] = 335
Probability of Lossfor Plan I = 12%
Consider two investment plans, designs, or operational decisionsConsider two investment plans, designs, or operational decisions
4
ConclusionsConclusions
Risk can only be assessed after a plan has been selected but it cannot be Risk can only be assessed after a plan has been selected but it cannot be
managed during the optimization stage (even when stochastic optimization managed during the optimization stage (even when stochastic optimization
including uncertainty has been performed). including uncertainty has been performed).
The decision maker has two simultaneous objectives:The decision maker has two simultaneous objectives:
There is a need to develop new models that allow not only assessing but managingThere is a need to develop new models that allow not only assessing but managing
financial risk. financial risk.
• Maximize Expected Profit. Maximize Expected Profit.
• Minimize Risk ExposureMinimize Risk Exposure
5
What does Risk Management mean?What does Risk Management mean?
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
0.20
-300 -200 -100 0 100 200 300 400 500 600 700 800Profit x
Probability
REDUCE THESE FREQUENCIES
OR…
INCREASE THESE FREQUENCIES
One wants to modify the profit distribution in order to satisfy One wants to modify the profit distribution in order to satisfy the preferences of the decision makerthe preferences of the decision maker
OR BOTH!!!!
6
Characteristics of Two-StageCharacteristics of Two-StageStochastic Optimization Models Stochastic Optimization Models
PhilosophyPhilosophy• Maximize the Maximize the Expected ValueExpected Value of the objective over all possible realizations of of the objective over all possible realizations of uncertain parameters.uncertain parameters.• Typically, the objective is Typically, the objective is Expected ProfitExpected Profit , usually , usually Net Present ValueNet Present Value..• Sometimes the minimization of Sometimes the minimization of CostCost is an alternative objective. is an alternative objective.
UncertaintyUncertainty• Typically, the uncertain parameters are: Typically, the uncertain parameters are: market demands, availabilities,market demands, availabilities, prices, process yields, rate of interest, inflation, etc.prices, process yields, rate of interest, inflation, etc.• In Two-Stage Programming, uncertainty is modeled through a finite numberIn Two-Stage Programming, uncertainty is modeled through a finite number of independent of independent ScenariosScenarios..• Scenarios are typically formed by Scenarios are typically formed by random samplesrandom samples taken from the probability taken from the probability distributions of the uncertain parameters.distributions of the uncertain parameters.
7
First-Stage DecisionsFirst-Stage Decisions• Taken before the uncertainty is revealed. They usually correspond to structural Taken before the uncertainty is revealed. They usually correspond to structural decisions (not operational). decisions (not operational). • Also called “Here and Now” decisions.Also called “Here and Now” decisions.• Represented by “Design” Variables.Represented by “Design” Variables.• Examples:Examples:
Characteristics of Two-StageCharacteristics of Two-StageStochastic Optimization Models Stochastic Optimization Models
−To build a plant or not. How much capacity should be added, etc. To build a plant or not. How much capacity should be added, etc. −To place an order now. To place an order now. −To sign contracts or buy options. To sign contracts or buy options. −To pick a reactor volume, to pick a certain number of trays and size To pick a reactor volume, to pick a certain number of trays and size the condenser and the reboiler of a column, etc the condenser and the reboiler of a column, etc
8
Second-Stage DecisionsSecond-Stage Decisions• Taken in order to adapt the plan or design to the uncertain parameters Taken in order to adapt the plan or design to the uncertain parameters realization.realization.• Also called “Recourse” decisions.Also called “Recourse” decisions.• Represented by “Control” Variables.Represented by “Control” Variables.• Example: the operating level; the production slate of a plant.Example: the operating level; the production slate of a plant.
• Sometimes first stage decisions can be treated as second stage decisions. Sometimes first stage decisions can be treated as second stage decisions. In such case the problem is called a multiple stage problem. In such case the problem is called a multiple stage problem.
ShortcomingsShortcomings• The model is unable to perform risk management decisions.The model is unable to perform risk management decisions.
Characteristics of Two-StageCharacteristics of Two-StageStochastic Optimization Models Stochastic Optimization Models
9
Two-Stage Stochastic FormulationTwo-Stage Stochastic Formulation
LINEAR LINEAR MODEL SPMODEL SP
xcyqpMax T
ss
Tss
bAx
Xxx 0sss hWyxT
0sy
s.t.
First-Stage ConstraintsFirst-Stage Constraints
Second-Stage ConstraintsSecond-Stage Constraints
RecourseRecourseFunctionFunction
First-StageFirst-StageCostCost
First stage variables
Second Stage Variables
Technology matrix
Recourse matrix (Fixed Recourse)
Sometimes not fixed (Interest rates in Portfolio Optimization)
Complete recourse: the recourse cost (or profit) for every possible uncertainty realization remains finite, independently of the first-stage decisions (x).
Relatively complete recourse: the recourse cost (or profit) is feasible for the set of feasible first-stage decisions. This condition means that for every feasible first-stage decision, there is a way of adapting the plan to the realization of uncertain parameters.
We also have found that one can sacrifice efficiency for certain scenarios to improve risk management. We do not know how to call this yet.
Let us leave it linear because as is it is complex enough.!!!
10
Robust Optimization Using Variance (Robust Optimization Using Variance (Mulvey et al., 1995)Mulvey et al., 1995)
Previous Approaches to Risk ManagementPrevious Approaches to Risk Management
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
-2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5Profit
Profit PDF Expected Profit
Variance is a measurefor the dispersionof the distribution
Desirable Penalty
Maximize E[Profit] - Maximize E[Profit] - ·V[Profit]·V[Profit]
Underlying Assumption: Underlying Assumption: Risk is monotonic with variabilityRisk is monotonic with variability
Undesirable Penalty
11
Robust Optimization Using VarianceRobust Optimization Using Variance
DrawbacksDrawbacks
• Variance is a symmetric risk measure: profits both above and below the targetVariance is a symmetric risk measure: profits both above and below the targetlevel are penalized equally. We only want to penalize profits below the target.level are penalized equally. We only want to penalize profits below the target.
• Introduces non-linearities in the model, which results in serious computationalIntroduces non-linearities in the model, which results in serious computationaldifficulties, specially difficulties, specially in large-scale problems.in large-scale problems.
• The model may render solutions that are stochastically dominated by others.The model may render solutions that are stochastically dominated by others.This is known in the literature as not showing Pareto-Optimality. In other wordsThis is known in the literature as not showing Pareto-Optimality. In other words
there is a better solution (ythere is a better solution (yss,x,x**) than the one obtained ) than the one obtained (y(yss**,x*). ,x*).
12
Robust Optimization using Upper Partial Mean (Ahmed and Sahinidis, 1998)Robust Optimization using Upper Partial Mean (Ahmed and Sahinidis, 1998)
Previous Approaches to Risk ManagementPrevious Approaches to Risk Management
x
0.0
0.1
0.2
0.3
0.4
-2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5Profit x
Profit PDF
UPM = E[x ]
E[x ]
UPM = 0.50 UPM = 0.44
Maximize E[Profit] - Maximize E[Profit] - ·UPM·UPM
Underlying Assumption: Underlying Assumption: Risk is monotonic with lower variabilityRisk is monotonic with lower variability
13
Robust Optimization using the UPMRobust Optimization using the UPM
AdvantagesAdvantages
• Linear measureLinear measure
Robust Optimization using the UPMRobust Optimization using the UPM
DisadvantagesDisadvantages
• The UPM may misleadingly favor non-optimalThe UPM may misleadingly favor non-optimal second-stage decisions.second-stage decisions.• Consequently, financial risk is not managed properly and solutions with higher riskConsequently, financial risk is not managed properly and solutions with higher risk
than the one obtained using the traditional two-stage formulation may be obtained.than the one obtained using the traditional two-stage formulation may be obtained.
• The model losses its scenario-decomposable structure and stochastic decompositionThe model losses its scenario-decomposable structure and stochastic decompositionmethods can no longer be used to solve it.methods can no longer be used to solve it.
14
Robust Optimization using the UPMRobust Optimization using the UPM
s
Skkks ProfitProfitpMax ; 0
= 3= 3Profits s
Case I Case II Case I Case II
S1 150 100 0 0
S2 125 100 0 0
S3 75 75 25 6.25
S4 50 50 50 31.25
E[Profit] 100.00 81.25
UPM 18.75 9.38
Objective 43.75 53.13
Ss
sspUPM
Objective Function: Maximize E[Profit] - Objective Function: Maximize E[Profit] - ·UPM·UPM
Downside scenarios are the same, but the UPM is affected by Downside scenarios are the same, but the UPM is affected by the change in expected profit due to a different upside distribution. the change in expected profit due to a different upside distribution. As a result a wrong choice is made. As a result a wrong choice is made.
15
Effect of Non-Optimal Second-Stage DecisionsEffect of Non-Optimal Second-Stage Decisions
Robust Optimization using the UPMRobust Optimization using the UPM
P1 A
P2 B
P1 A
P2 B
P1 A
P2 B
-380
-360
-340
-320
-300
-280
-260
-240
-220
-200
0 20 40 60 80 100 120 140 160 180 200 220
E[Profit ]
Robustness Solution
Robustness Solution with Optimal Second-Stage Decisions
0
2
4
6
8
10
12
14
16
18
0 20 40 60 80 100 120 140 160 180 200 220
UPM
Robustness Solution
Robustness Solution with Optimal Second-Stage Decisions
Both technologies are able to produce two products with different production cost and at different yield per unit of installed capacity
-380
-360
-340
-320
-300
-280
-260
-240
-220
-200
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17UPM
E[Profit ]
Robustness Solution
Robustness Solution with Optimal Second-Stage Decisions
16
OTHER APPROACHESOTHER APPROACHESCheng, Subrahmanian and Westerberg (2002, unpublished)Cheng, Subrahmanian and Westerberg (2002, unpublished)
This paper proposes a new design paradigm of which risk is just one component. This paper proposes a new design paradigm of which risk is just one component. We will revisit this issue later in the talk.We will revisit this issue later in the talk.
− Multiobjective Approach: Considers Downside Risk, ENPV and Process Multiobjective Approach: Considers Downside Risk, ENPV and Process Life Cycle as alternative Objectives.Life Cycle as alternative Objectives.− Multiperiod Decision process modeled as a Markov decision process Multiperiod Decision process modeled as a Markov decision process with recourse.with recourse.− The problem is sometimes amenable to be reformulated as a sequence The problem is sometimes amenable to be reformulated as a sequence of single-period sub-problems, each being a two-stage stochastic program of single-period sub-problems, each being a two-stage stochastic program with recourse. These can often be solved backwards in time to obtain with recourse. These can often be solved backwards in time to obtain Pareto Optimal solutions. Pareto Optimal solutions.
17
OTHER APPROACHESOTHER APPROACHES
Risk Premium (Applequist, Pekny and Reklaitis, 2000)Risk Premium (Applequist, Pekny and Reklaitis, 2000)
− Observation: Rate of return varies linearly with variability. The Observation: Rate of return varies linearly with variability. The of such dependance is called Risk Premium. of such dependance is called Risk Premium. − They suggest to benchmark new investments against the historical They suggest to benchmark new investments against the historical − risk premium by using a two objective (risk premium and profit) risk premium by using a two objective (risk premium and profit) − problem. problem. −The technique relies on using variance as a measure of variability.The technique relies on using variance as a measure of variability.
18
ConclusionsConclusions
• The minimization of Variance penalizes both sides of the mean. The minimization of Variance penalizes both sides of the mean. • The Robust Optimization Approach using Variance or UPM is not suitable The Robust Optimization Approach using Variance or UPM is not suitable for risk management.for risk management.• The Risk Premium Approach (Applequist et al.) has the same problems The Risk Premium Approach (Applequist et al.) has the same problems as the penalization of variance.as the penalization of variance.
THUS, THUS, • Risk should be properly defined and Risk should be properly defined and directly directly incorporated in the models to incorporated in the models to manage it. manage it. • The multiobjective Markov decision process (Applequist et al, 2000) The multiobjective Markov decision process (Applequist et al, 2000) is very closely related to ours and can be considered complementary. In is very closely related to ours and can be considered complementary. In fact (Westerberg dixit) it can be extended to match ours in the definition fact (Westerberg dixit) it can be extended to match ours in the definition of risk and its multilevel parametrization. of risk and its multilevel parametrization.
Previous Approaches to Risk ManagementPrevious Approaches to Risk Management
19
Financial Risk = Probability that a plan or Financial Risk = Probability that a plan or design does not meet a certain profit targetdesign does not meet a certain profit target
Probabilistic Definition of RiskProbabilistic Definition of Risk
zzss is a new is a new binarybinary variable variable
Formal Definition of Financial RiskFormal Definition of Financial Risk
ProfitPxRisk ),(
Scenarios are independent eventsScenarios are independent events s
ss ProfitPpxRisk ),(
else0
If1 ss
ProfitProfitP
ss zProfitP
s
ss zpxRisk ),(
For each scenario the profit is eitherFor each scenario the profit is eithergreater/equal or smaller than the targetgreater/equal or smaller than the target
20
Financial Risk InterpretationFinancial Risk Interpretation
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
0.20
Profit x
Probability
Cumulative Probability = Risk (x ,)
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
Profit x
Area = Risk (x ,)
x fixed
Profit PDF f (x )
21
Cumulative Risk CurveCumulative Risk Curve
Our intention is to modify the shape and location of thiscurve according to the attitude towards risk of the decision maker
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250Profit (M$)
Risk
22
Risk Preferences and Risk CurvesRisk Preferences and Risk Curves
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0Profit
Risk
Risk-AverseInvestor's ChoiceE[Profit ] = 0.4
Risk-T akerInvestor's ChoiceE[Profit ] = 1.0
23
Risk Curve PropertiesRisk Curve PropertiesA plan or design with Maximum E[A plan or design with Maximum E[ProfitProfit] (i.e. optimal in Model SP) sets a ] (i.e. optimal in Model SP) sets a theoretical limit for financial risk: it is impossible to find a feasible plan/design theoretical limit for financial risk: it is impossible to find a feasible plan/design having a risk curve entirely beneath this curve.having a risk curve entirely beneath this curve.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Profit
Risk
MaximumE[Profit ]
Impossiblecurve
Possiblecurve
24
Minimizing Risk: a Multi-Objective ProblemMinimizing Risk: a Multi-Objective Problem
)1,0()1(
s
ssT
sTs
ssT
sTs
zzUxcyq
zUxcyq
Xxx 0
0sy
s.t.
xcyqp ProfitMax E T
ssss
s
ss zp Min Risk 11
s
sisi zp Min Risk
...
bAx
sss hWyxT
Multiple Objectives:Multiple Objectives:• At each profit we want minimize the associated riskAt each profit we want minimize the associated risk• We also want to maximize the expected profitWe also want to maximize the expected profit
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Target Profit
Risk
x fixed
1 2 3 4
Min Risk (x ,1)
Min Risk (x ,2)
Min Risk (x ,3)
Min Risk (x ,4)
Max E[Profit (x )]
25
Restricted RiskRestricted Risk MODELMODEL
Risk ManagementRisk ManagementConstraintsConstraints
xcyqpMax T
ssss
s.t.
is
sis εzp
)1,0()1(
s
sisiT
sTs
sisiT
sTs
zzUxcyq
zUxcyq
Xxx 0
0sy
bAx
sss hWyxT
Forces Risk to be lowerForces Risk to be lowerthan a specified levelthan a specified level
Parametric Representations of theParametric Representations of the Multi-Objective Model – Restricted RiskMulti-Objective Model – Restricted Risk
26
Parametric Representations of theParametric Representations of the Multi-Objective Model – Penalty for RiskMulti-Objective Model – Penalty for Risk
s.t.
i s
sisiT
ss
Tss zpxc yqpMax
Penalty TermPenalty Term
Risk PenaltyRisk Penalty MODELMODEL
Risk ManagementRisk ManagementConstraintsConstraints
)1,0()1(
s
sisiT
sTs
sisiT
sTs
zzUxcyq
zUxcyq
Xxx 0
0sy
bAx
sss hWyxT
Define several profit Define several profit Targets and penaltyTargets and penaltyweights to solve theweights to solve themodel using a multi-model using a multi-parametric approachparametric approach
STRATEGYSTRATEGY
27
AdvantagesAdvantages
• Risk can be effectively managed according to the decision maker’s criteria.Risk can be effectively managed according to the decision maker’s criteria.
• The models can adapt to risk-averse or risk-taker decision makers, and theirThe models can adapt to risk-averse or risk-taker decision makers, and their risk preferences are easily matched using the risk curves.risk preferences are easily matched using the risk curves.
• A full spectrum of solutions is obtained. These solutions always haveA full spectrum of solutions is obtained. These solutions always haveoptimal second-stage decisions.optimal second-stage decisions.
• Model Risk Penalty conserves all the properties of the standard two-stageModel Risk Penalty conserves all the properties of the standard two-stage stochastic formulation.stochastic formulation.
Risk Management using the New ModelsRisk Management using the New Models
DisadvantagesDisadvantages
• The use of binary variables is required, which increases the computational The use of binary variables is required, which increases the computational time to get a solution. This is a major limitation for large-scale problems.time to get a solution. This is a major limitation for large-scale problems.
28
Computational IssuesComputational Issues
Risk Management using the New ModelsRisk Management using the New Models
• The most efficient methods to solve stochastic optimization problems reportedThe most efficient methods to solve stochastic optimization problems reportedin the literature exploit the decomposable structure of the model. in the literature exploit the decomposable structure of the model.
• This property means that each scenario defines an independent second-stageThis property means that each scenario defines an independent second-stageproblem that can be solved separately from the other scenarios once the first-problem that can be solved separately from the other scenarios once the first-stage variables are fixed.stage variables are fixed.
• The Risk Penalty Model is decomposable whereas Model Restricted Risk is not.The Risk Penalty Model is decomposable whereas Model Restricted Risk is not.Thus, the first one is model is preferable.Thus, the first one is model is preferable.
• Even using decomposition methods, the presence of binary variables in bothEven using decomposition methods, the presence of binary variables in bothmodels constitutes a major computational limitation to solve large-scale problems.models constitutes a major computational limitation to solve large-scale problems.
• It would be more convenient to measure risk indirectly such that binary variablesIt would be more convenient to measure risk indirectly such that binary variablesin the second stage are avoided.in the second stage are avoided.
29
,, xExDRiskDownside Risk Downside Risk (Eppen et al, 1989)(Eppen et al, 1989) == Expected Value of the PositiveExpected Value of the Positive Profit Deviation from the targetProfit Deviation from the target
Downside RiskDownside Risk
Positive Profit Deviation fromPositive Profit Deviation fromTarget Target
Formal definition of Downside RiskFormal definition of Downside Risk
Otherwise0
If,
xProfitxProfitx
s
sspxDRisk ,
The Positive Profit Deviation isThe Positive Profit Deviation isalso defined for each scenarioalso defined for each scenario
Otherwise0
If sss
ProfitProfit
30
Downside Risk InterpretationDownside Risk Interpretation
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
Profit x
x fixed
DRisk (x ,) = E[(x ,)]
Profit PDF f (x )
ò
¥xxx dfxDRisk )(),(
31
Downside Risk & Probabilistic RiskDownside Risk & Probabilistic Risk
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Profit x
Risk (x ,x )
x fixed
Area = DRisk (x ,)
ò
¥xx dxRiskxDRisk ),(),(
32
Two-Stage Model using Downside RiskTwo-Stage Model using Downside Risk
s.t.
sss
T
ss
Tss pxc yqpMax
Penalty TermPenalty Term
MODEL DRiskMODEL DRisk
Downside Downside Risk ConstraintsRisk Constraints
)( xcyq Ts
Tss
Xxx 0
0sy
bAx
sss hWyxT
0s
AdvantagesAdvantages
• Same as models using RiskSame as models using Risk
• Does not require the use ofDoes not require the use ofbinary variablesbinary variables
• Potential benefits from thePotential benefits from theuse of decomposition methodsuse of decomposition methods
StrategyStrategy
Solve the model using differentSolve the model using different profit targets to get a full spectrumprofit targets to get a full spectrum of solutions. Use the risk curves toof solutions. Use the risk curves to select the solution that better suitsselect the solution that better suits the decision maker’s preferencethe decision maker’s preference
33
Two-Stage Model using Downside RiskTwo-Stage Model using Downside Risk
Warning: Warning: The same risk may imply different Downside Risks. The same risk may imply different Downside Risks.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0Profit
Risk
DRisk (Design I , 0.5) = 0.2 Risk (Design I , 0.5) = 0.5
DRisk (Design II , 0.5) = 0.2 Risk (Design II , 0.5) = 0.309
Immediate Consequence: Immediate Consequence: Minimizing downside risk does not guarantee minimizing risk.Minimizing downside risk does not guarantee minimizing risk.
34
Riskoptimizer (Palisades) and CrystalBall (Decisioneering)Riskoptimizer (Palisades) and CrystalBall (Decisioneering)
• Use excell modelsUse excell models• Allow uncertainty in a form of distributionAllow uncertainty in a form of distribution• Perform Montecarlo Simulations or use genetic algorithmsPerform Montecarlo Simulations or use genetic algorithms to optimize (Maximize ENPV, Minimize Variance, etc.) to optimize (Maximize ENPV, Minimize Variance, etc.)
Financial Software. Large varietyFinancial Software. Large variety
•Some use the concept of downside riskSome use the concept of downside risk
• In most of these software, Risk is mentioned but not manipulated directly.In most of these software, Risk is mentioned but not manipulated directly.
Commercial SoftwareCommercial Software
35
Process Planning Under UncertaintyProcess Planning Under Uncertainty
OBJECTIVESOBJECTIVES:: Maximize Expected Net Present ValueMaximize Expected Net Present Value
Minimize Financial RiskMinimize Financial Risk
Production LevelsProduction Levels
DETERMINEDETERMINE:: Network ExpansionsNetwork ExpansionsTimingTimingSizingSizingLocationLocation
GIVEN:GIVEN: Process NetworkProcess Network Set of ProcessesSet of ProcessesSet of ChemicalsSet of Chemicals
A 1C2
D3
B
Forecasted DataForecasted DataDemands & AvailabilitiesDemands & AvailabilitiesCosts & PricesCosts & PricesCapital BudgetCapital Budget
36
Process Planning Under UncertaintyProcess Planning Under Uncertainty
Design Variables: Design Variables: to be decided before the uncertainty revealsto be decided before the uncertainty reveals
x EitYit , , QitY: Decision of building process Y: Decision of building process ii in period in period ttE: Capacity expansion of process E: Capacity expansion of process ii in period in period ttQ: Total capacity of process Q: Total capacity of process ii in period in period tt
Control Variables:Control Variables: selected after the uncertain parameters become knownselected after the uncertain parameters become known
S: S: Sales of product Sales of product jj in market in market ll at time at time tt and scenario and scenario ss P: P: Purchase of raw mat. Purchase of raw mat. jj in market in market ll at time at time t t and scenario and scenario ssW: W: Operating level of of process Operating level of of process ii in period in period tt and scenario and scenario ss
ys PjltsSjlts , , Wits
37
ExampleExample
Uncertain Parameters: Uncertain Parameters: Demands, Availabilities, Sales Price, Purchase PriceDemands, Availabilities, Sales Price, Purchase Price
Total of 400 ScenariosTotal of 400 Scenarios
Project Staged in 3 Time Periods of 2, 2.5, 3.5 yearsProject Staged in 3 Time Periods of 2, 2.5, 3.5 years
Process 1Chemical 1 Process 2
Chemical 5
Chemical 2
Chemical 6
Process 3
Chemical 3
Process 5
Chemical 7
Chemical 8
Process 4
Chemical 4
38
Period 1Period 12 years2 yearsPeriod 2Period 22.5 years2.5 yearsPeriod 3Period 33.5 years3.5 years
Process 1Chemical 1
Chemical 5
Process 3
Chemical 3
Chemical 7
10.23 kton/yr
22.73 kton/yr
5.27 kton/yr
5.27 kton/yr
19.60 kton/yr
19.60 kton/yr
Process 1Chemical 1
Chemical 5
Process 3
Chemical 3
Process 5Chemical 7
Chemical 8
Process 4Chemical 4
10.23 kton/yr
22.73 kton/yr
22.73 kton/yr
22.73 kton/yr
4.71 kton/yr
4.71 kton/yr
41.75 kton/yr
20.87 kton/yr
20.87 kton/yr
20.87 kton/yr
Chemical 1 Process 2
Chemical 5
Chemical 2
Chemical 6
Process 3
Chemical 3
Process 5Chemical 7
Chemical 8
Process 4Chemical 4
22.73 kton/yr
22.73 kton/yr 22.73 ton/yr
80.77 kton/yr 80.77 kton/yr44.44 kton/yr
14.95 kton/yr
29.49 kton/yr
29.49 kton/yr
43.77 kton/yr
29.49 kton/yr
21.88 kton/yr
21.88 kton/yr
21.88 kton/yr
Process 1
Example – Solution with Max ENPVExample – Solution with Max ENPV
39
Period 1Period 12 years2 yearsPeriod 2Period 22.5 years2.5 yearsPeriod 3Period 33.5 years3.5 years
Process 1Chemical 1
Chemical 5
Process 3
Chemical 3
Chemical 7
10.85 kton/yr
22.37 kton/yr
5.59 kton/yr
5.59 kton/yr
19.30 kton/yr
19.30 kton/yr
Process 1Chemical 1
Chemical 5
Process 3
Chemical 3
Process 5Chemical 7
Chemical 8
Process 4Chemical 4
10.85 kton/yr
22.37 kton/yr
22.37 kton/yr
22.43 kton/yr
4.99 kton/yr
4.99 kton/yr
41.70 kton/yr
20.85 kton/yr
20.85 kton/yr
20.85 kton/yr
Process 1Chemical 1 Process 2
Chemical 5
Chemical 2
Chemical 6
Process 3
Chemical 3
Process 5Chemical 7
Chemical 8
Process 4Chemical 4
22.37 kton/yr
22.37 kton/yr 22.77 ton/yr
10.85 kton/yr 10.85 kton/yr7.54 kton/yr
2.39 kton/yr
5.15 kton/yr
5.15 kton/yr
43.54 kton/yr
5.15 kton/yr
21.77 kton/yr
21.77 kton/yr
21.77 kton/yr
Same final structure, different production capacities. Same final structure, different production capacities.
Example – Solution with Min DRisk(Example – Solution with Min DRisk(=900)=900)
40
Example – Solution with Max ENPVExample – Solution with Max ENPV
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250NPV (M$)
Risk
PP solut ion
E[NPV ] = 1140 M$
41
Example – Risk Management SolutionsExample – Risk Management Solutions
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250NPV (M$)
Risk
P P500600700800900100011001200130014001500
increases
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250NPV (M$)
Risk
= 900ENPV = 908 = 1100
ENPV = 1074
PPENPV =1140
0.0000
0.0002
0.0004
0.0006
0.0008
0.0010
0.0012
0.0014
0.0016
0.0018
0.0020
0.0022
0.0024
0.0026
0 500 1000 1500 2000 2500 3000NPV ( x , M$ )
NPV PDF f (x)
= 900
= 1100PP
42
Process Planning with InventoryProcess Planning with Inventory
OBJECTIVESOBJECTIVES:: Maximize Expected Net Present ValueMaximize Expected Net Present Value
Minimize Financial RiskMinimize Financial Risk
The mass balance is modified such that now a certain levelThe mass balance is modified such that now a certain levelof inventory for raw materials and products is allowedof inventory for raw materials and products is allowed
A storage cost is included in the objectiveA storage cost is included in the objective
PROBLEM DESCRIPTION:PROBLEM DESCRIPTION:
A1
2
D3
B D
MODELMODEL::
43
Period 1Period 12 years2 yearsPeriod 2Period 22.5 years2.5 yearsPeriod 3Period 33.5 years3.5 years Chemical 5
Chemical 2
Chemical 6
51.95 kton/yr 22.36 kton/yr
Process 1 Process 2
5.14 kton/yr
12.48 kton/yr 1.05
kton/yr
16.28 kton/yr
Chemical 1
33.90kton/yr
2.88 kton/yr
11.67 kton/yr
0.81 kton/yr
12.48 kton/yr
4.77 kton/yr
Chemical 6
Process 3
36.45 kton/yr
51.95 kton/yr 76.81 kton/yr
Process 1
1.62kton
Chemical 5
10.28kton
11.80 kton/yr
Chemical 2
2.11kton
27.24 kton/yr 0.60
kton/yr
Chemical 74.65
kton/yr31.09
kton/yr
Chemical 1
Chemical 3
39.04kton/yr
35.74kton/yr
5.75kton
0.42 kton/yr
26.34 kton/yr
0.90 kton/yr
27.24 kton/yr
Process 2
1.18 kton/yr
Chemical 1Chemical 6
Process 3
Chemical 3
Process 5Chemical 8
Process 4 Chemical 4
26.77 kton/yr
36.45 kton/yr 26.77 kton/yr
76.81 kton/yr 76.81 kton/yr
43.14kton/yr
25.41kton/yr
Process 1 Process 2
3.86kton
Chemical 7
11.64kton
25.41 kton/yr
Chemical 5
7.32kton
13.61 kton/yr
Chemical 2
3.86kton
30.44 kton/yr
3.29 kton/yr
0.04 kton/yr
6.80kton 1.94
kton/yr
11.91kton 3.40
kton/yr
44.13 kton/yr
2.09 kton/yr
31.47 kton/yr
22.12 kton/yr
1.10 kton/yr
31.47 kton/yr
1.03 kton/yr
Example with Inventory – SP SolutionExample with Inventory – SP Solution
44
Example with InventoryExample with InventorySolution with Min DRisk (Solution with Min DRisk (=900)=900)
3.64 kton/yr
Process 3
22.15 kton/yr
11.23 kton/yr
Process 1
Chemical 55.80 kton/yr
Chemical 73.69
kton/yr18.46
kton/yr
Chemical 1
Chemical 3
6.63kton/yr
25.79kton/yr
0.51 kton/yr
0.32 kton/yr
Chemical 1
Process 3
Chemical 3
Process 5 Chemical 8
Process 4Chemical 4
23.38 kton/yr
22.15 kton/yr 23.38 kton/yr
11.23 kton/yr
5.73kton/yr
1.64kton/yr
Process 1
Chemical 7
7.38kton
22.18 kton/yr
Chemical 5
0.64kton
5.61 kton/yr
1.60 kton/yr
1.01kton
0.02 kton/yr
7.27kton 1.29
kton/yr
41.68 kton/yr 20.58
kton/yr
0.20 kton/yr
0.10 kton/yr
20.54 kton/yr
Chemical 1Chemical 6
Process 3
Chemical 3
Process 5
Process 4
23.38 kton/yr
22.15 kton/yr 23.38 kton/yr
11.23 kton/yr 11.23 kton/yr
7.48kton/yr
Process 1 Process 2
Chemical 7
3.37kton
22.85 kton/yr
Chemical 5
0.90kton
2.39 kton/yr
Chemical 2
5.39 kton/yr
0.96 kton/yr
1.07kton 0.30
kton/yr
4.05kton 1.16
kton/yr
43.72 kton/yr
0.26 kton/yr
5.39 kton/yr
22.04 kton/yr
5.39 kton/yr
Chemical 4 0.51kton
Chemical 8
1.17kton/yr
4.11kton
23.00 kton/yr
0.15 kton/yr
Period 1Period 12 years2 yearsPeriod 2Period 22.5 years2.5 yearsPeriod 3Period 33.5 years3.5 years
45
Example with Inventory - SolutionsExample with Inventory - Solutions
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250 3500 3750 4000NPV (M$)
Risk
PP solut ion
E[NPV ] = 1140 M$PPI solut ion
E[NPV ] = 1237 M$
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
0 250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250 3500 3750 4000NPV (M$)
Risk
= 900ENPV = 980
= 1400ENPV = 1184
PPIENPV = 1140
PPIENPV = 1237With Inventory
WithoutInventory
DRisk
DRisk
46
Downside Expected ProfitDownside Expected ProfitDefinition: Definition:
0
125
250
375
500
625
750
875
1000
1125
1250
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0Risk
CEP (M$)
PP solution
E[NPV ] = 1140 M$
= 900
E[NPV ] = 908 M$
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
250 500 750 1000 1250 1500 1750 2000 2250 2500 2750 3000 3250NPV (M$)
Risk
= 900 = 1100
PP
Up to 50% of risk (confidence?) the lower ENPV solution has higher profit Up to 50% of risk (confidence?) the lower ENPV solution has higher profit expectations. expectations.
),(),(),(),( ò
¥ xDRiskxRiskdxfpxDENPV xxx
47
Value at RiskValue at RiskDefinition: Definition:
VaR=zVaR=zpp for symmetric distributions (Portfolio optimization) for symmetric distributions (Portfolio optimization)
VaR is given by the difference between the mean value of the profit and the profit value corresponding to the p-quantile.
),()]([),( 1 xRiskxProfitEpxVaR
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
Profit x
Area = Risk (x ,)
x fixed
Profit PDF
)]([ xProfitE
),( pxVaR
48
COMPUTATIONAL APPROACHESCOMPUTATIONAL APPROACHES
Sampling Average Approximation MethodSampling Average Approximation Method::
−Solve M times the problem using only N scenarios. Solve M times the problem using only N scenarios. −If multiple solutions are obtained, use the first stage variables to solve the If multiple solutions are obtained, use the first stage variables to solve the problem with a large number of scenarios N’>>N to determine the optimum. problem with a large number of scenarios N’>>N to determine the optimum.
− First Stage variables are complicating variables. First Stage variables are complicating variables. − This leaves a primal over second stage variables, which is decomposable. This leaves a primal over second stage variables, which is decomposable.
Generalized Benders Decomposition Algorithm Generalized Benders Decomposition Algorithm (Benders Here)(Benders Here)::
49
ConclusionsConclusionsA probabilistic definition of Financial Risk has been introduced in the A probabilistic definition of Financial Risk has been introduced in the framework of two-stage stochastic programming. Theoretical properties offramework of two-stage stochastic programming. Theoretical properties ofrelated to this definition were explored.related to this definition were explored.
Using downside risk leads to a model that is decomposable in scenarios and thatUsing downside risk leads to a model that is decomposable in scenarios and thatallows the use of efficient solution algorithms. For this reason, it is suggested allows the use of efficient solution algorithms. For this reason, it is suggested that this model be used to manage financial risk.that this model be used to manage financial risk.
New formulations capable of managing financial risk have been introduced.New formulations capable of managing financial risk have been introduced.The multi-objective nature of the models allows the decision maker to chooseThe multi-objective nature of the models allows the decision maker to choosesolutions according to his risk policy. The cumulative risk curve is used as asolutions according to his risk policy. The cumulative risk curve is used as atool for this purpose.tool for this purpose.
To overcome the mentioned computational difficulties, the concept of DownsideTo overcome the mentioned computational difficulties, the concept of DownsideRisk was examined, finding that there is a close relationship between thisRisk was examined, finding that there is a close relationship between thismeasure and the probabilistic definition of risk.measure and the probabilistic definition of risk.
The models using the risk definition explicitly require second-stage binary The models using the risk definition explicitly require second-stage binary variables. This is a major limitation from a computational standpoint.variables. This is a major limitation from a computational standpoint.
An example illustrated the performance of the models, showing how the riskAn example illustrated the performance of the models, showing how the riskcurves can be changed in relation to the solution with maximum expected profit.curves can be changed in relation to the solution with maximum expected profit.