+ All Categories
Home > Documents > Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review...

Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review...

Date post: 31-Mar-2015
Category:
Upload: alana-fildes
View: 236 times
Download: 2 times
Share this document with a friend
Popular Tags:
28
Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review Constrained Optimization
Transcript
Page 1: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Topic 03: Optimal Decision Making

• Overview

• Unconstrained Optimization

• Univariate Calculus – review

• Constrained Optimization

• Multivariate Calculus - review

• Constrained Optimization

Page 2: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Overview

• Many economic decisions involve trying to decide what is the “best” decision to make.

• These problems involving trying to optimize some objective, such as maximize profits, or minimize risk exposure, etc.

Page 3: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Analytical Approach to Optimal Decision Making

• Express the relationship between the objective and various decision and exogenous variables.

• Dependent variable =f(independent variables)

• Example: Profit = f(quantity) or = f(Q)» Illustration: Find the quantity of output that

maximizes profit: (Q) = 16·Q - Q2

Page 4: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Unconstrained Optimization

• Unconstrained Optimization involves finding the optimum to some decision problem in which there are no constraints.

Page 5: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Average Profit = Profit / Q

• Slope of ray from the origin» Rise / Run» Profit / Q = average

profit

• Maximizing average profit doesn’t maximize total profit

MAX

C

B

profits

Q

PROFITS

quantity

Page 6: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Marginal Profits = Q

• profits of the last unit produced

• maximum marginal profits occur at the inflection point (A)

• Decision Rule: produce where marginal profits = 0.

profits max

A

marginalprofits

Q

Q

averageprofits

averageprofits

BC

Page 7: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Optimization using calculus

• Calculus uses derivatives: • dP/dQ = lim P / Q as Q goes

to 0.

• Decision rule - To maximize profits, find where dP/dQ = 0 -- first order condition

Page 8: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Review of Differentiation Rules

• Constant Y = c dY/dX = 0 Y = 5

dY/dX = 0

• Line Y = c•X dY/dX = c Y = 5•X

dY/dX = 5

• Power Y = cXb dY/dX = b•c•X b-1 Y = 5•X2

dY/dX = 10•X

Name Function Derivative Example

Page 9: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

• Sum Rule: If Y = G(X) + H(X) , thendY/dX = dG/dX + dH/dX

Example: Y = 5•X + 5•X2 dY/dX = 5 + 10•X

• Product Rule: If Y = G(X)•H(X), then

dY/dX = (dG/dX)H + (dH/dX)G

Example: Y = (5•X)(5•X2 )

dY/dX = 5(5•X2 ) + (10•X)(5•X) = 75•X2

Review of Differentiation Rules

Page 10: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

•Quotient Rule: If Y = G(X) / H(X) ,

•Then dY/dX = (dG/dX)•H - (dH/dX)•G H2

Example: Y = (5•X) / (5•X2)

dY/dX = 5(5•X2) -(10•X)(5•X)

(5•X2)2

= -25X2 / 25X4 = -X-2

Review of Differentiation Rules

Page 11: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Review of Differentiation Rules

• Chain Rule:

• If Y = G [ H(X) ] , then dY/dX = (dG/dH)•(dH/dX)

Example: Y=(5+5X)2 ,

dY/dX = 2(5 + 5•X)1(5) = 50 + 50•X

Page 12: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Example #1: Maximization problem

• A profit function might look like an arch, rising to a peak and then declining at even larger outputs. A firm might sell huge amounts at very low prices, but discover that profits are low or negative.

• At the maximum, the slope of the profit function is zero. The first order condition for a maximum is that the derivative at that point is zero.

Page 13: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Example #1 (continued)

• Competitive Firm: Maximize Profits • where = TR - TC = P•Q - TC(Q)• Use our first order condition:

d/dQ = P - dTC/dQ = 0 • Decision Rule: P = MC

a function of Q

• Max P = 100•Q - Q2

100 -2•Q = 0 implies Q = 50 and P = 2,500

Page 14: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Static Optimal Decisions

• The scale of a project should expand until

• MB = MC

• Example: screening for prostate or breat cancer

• Example of marginal reasoning

MB

MC

frequency per decade

Page 15: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Capital Budgeting Example

• Investments• Project A: $500m 23.0%• Project B: $75m 18.0%• Project C: $50m 21.0%• Project D: $125m 16.0%• Project E: $300m 14.0%• Project F: $150m 13.0%• Project G: $250m 19.0%

Funds:First 250m 14.0%Next 250m 15.5%Next 100m 16.0%Next 250m 16.5%Next 200m 18.0%Next 200m 21.0%

Page 16: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

How Much to Produce ?

• Graph of output and profit

• Possible Rule:» Expand output until

profits turn down» But problem of

local maxima vs. global maximum

quantity B

MAX

GLOBALMAX

profit

A

Page 17: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Optimum Can Be Highest or Lowest

• Finding the maximum flying range for the Stealth Bomber is an optimization problem.

• Calculus teaches that when the first derivative is zero, the solution is at an optimum.

• The original Stealth Bomber study showed that a controversial flying V-wing design optimized the bomber's range, but the original researchers failed to find that their solution in fact minimized the range.

• It is critical that managers make decision that maximize, not minimize, profit potential!

Page 18: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Second Order Condition: One variable

• If the second derivative is negative, then it’s a maximum

• If the second derivative is positive, then it’s a minimum

Max = 100•Q - Q2

100 -2•Q = 0

second derivative is: -2 implies Q =50 is a MAX

Max= 50 + 5•X2

10•X = 0

second derivative is: 10 implies Q = 0 is a MIN

Page 19: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Multivariate Calculus• Economic relationships usually involve several

independent variables.• A partial derivative is like a controlled experiment

-- it holds the “other” variables constant For example, suppose that we are concerned with how demand for a product changes as its price changes, but assume that disposable income stays the same: Q = f (P, I ). Consequently, we are interested in: Q/P holds income constant

• We can often find partial derivatives using the above differentiation rules.

Page 20: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Example #2

• Sales are a function of advertising in newspapers and magazines ( X, Y)

• Max S = 200X + 100Y -10X2 -20Y2 +20XY• Differentiate with respect to X and Y and set equal

to zero.

S/X = 200 - 20X + 20Y= 0

S/Y = 100 - 40Y + 20X = 0

• solve for X & Y and Sales

Page 21: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Example #2 (continued)

• 200 - 20X + 20Y= 0• 100 - 40Y + 20X = 0• Adding them, the -20X and +20X cancel, so

we get 300 - 20Y = 0, or Y =15• Plug into one of them:

200 - 20X + 300 = 0, hence X = 25• To find Sales, plug into equation:

S = 200X + 100Y -10X2 -20Y2 +20XY = 3,250

Page 22: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Constrained Optimization

• Constrained Optimization involves finding the optimum to some decision problem in which the decision-maker faces constraints. » Examples: constraints of money, time, capacity, or energy.

• Sometimes managers know that some constraints are binding, which means that they are equality constraints.

• Sometimes managers face optimization problems in which the constraints are not necessarily binding. For example, you do not have to spend all your income. These constraints are inequality constraints. Linear programming and other optimization techniques can be used in these cases.

Page 23: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Format of Constrained Optimization

•Economic problems require tradeoffs forced on

us by the limits of our money, time, and energy.•Optimization involves an objective function

and one or more constraints.

• Maximize y = f(x1 , x2 , ..., xn )

Subject to g(x1 , x2 , ..., xn ) < b

• Minimize y = f(x1 , x2 , ..., xn )

Subject to g(x1 , x2 , ..., xn ) > b

Page 24: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Optimization with equality constraints

• An equality constraint is binding constraint that must be satisfied.

• An artificial variable is created for each constraint in the Lagrangian multiplier technique. This artificial variable is traditionally called lambda, .» Max L = (objective fct.) -{constraint set to zero}» Min L = (objective fct.) +{constraint set to zero}

• A binding constraint tends to reduce the solution set of an optimal decision making problem, and thereby constrains the optimal decision.

Page 25: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Example #3• Bounds Inc has determined through regression analysis

that its sales (S) are a function of the amount of advertising (measured in units) in two different media. This is given by the following relationship (X=newspapers, Y=magazines):

• S(X,Y) = 200X + 100Y - 10X2 - 20Y2 + 20XY

a. Find the level of newspaper and magazine advertising that maximizes the firm’s sales.

b. Calculate the firm’s sales at the optimal levels of newspaper and magazine advertising.

Page 26: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Example #3 (continued)

• Earlier we saw that: S/X = 200-20X+20Y =0S/Y = 100-40X+20Y =0

• You get X* = 15 units and Y* = 25 units.

• Plugging these optimal advertising levels back into the sales equation gives:

• S* = 200(25) + 100(15) – 10(25)2 – 20(15)2 + 20(25)(15) = 3250

Page 27: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Example #4

• Suppose that Bounds Inc. (in example #3) faces a budget constraint such that it can only purchase 20 units of advertising.

• A. Determine the level of newspaper and magazine advertising that maximizes sales given this constraint.

• B. What does the Lagrange multipler represent in this case?

• C. Compare the constrained optimization solution to the unconstrained optimization solution.

Page 28: Topic 03: Optimal Decision Making Overview Unconstrained Optimization Univariate Calculus – review Constrained Optimization Multivariate Calculus - review.

Example #4 (continued)•Max S = 200X + 100Y – 10X2 – 20Y2 +20XY subject to: X+Y=20.

•L= 200X + 100Y – 10X2 – 20Y2 +20XY – (X+Y-20)

•Take partial derivatives wrt X, Y, and :L/X = 200-20X+20Y – lL/Y = 100-40Y+20X – lL/ = -X - Y + 20

•Set the partial derivatives equal to 0 and solve for X, Y, and l. You will get: Y*=7 units, X* = 13 units, and l* = 80.

•The Lagrange multiplier represents the amount that sales would increase if we increased our advertising budget by 1 unit.

•Unconstrained solutions: X*=25, Y*=15, S*=3250

•Constrained solutions: X*=13, Y*=7, S*=2450


Recommended