UNC-Wilmington ECN 321
Department of Economics and Finance Dr. Chris Dumas
Non-linear Programming with Constraints, Lagranges Method
-- Example Problems
Example (1)
* Nonlinear programming problem
* One choice variable
* One constraint
Note: When we have a nonlinear problem with one or more constraints,
we need to use Lagrange's method.
max U = 2*ln(3X)
X
subject to: 5X 100
Because this is a non-linear programming problem with a constraint, we will use Lagrange's method as the
solution method. Converting the problem into an equivalent Lagrangian problem, we re-write the objective
function as a Lagrangian expression, denoted "L." In doing so, we incorporate the constraint into the
Lagrangian expression by introducing the new choice variable into the problem as a Lagrangian
multiplier:
max L = 2*ln(3X) + (100 - 5X)
X,
subject to: nothing
F.O.C.'s:
L 1
(1) 2( ) 3 5 0
X 3X
L
(2) 100 5X 0
The FOC equations provide us with two equations in two unknowns, X and . Solve these two equations
for the solution values of X and . There are several ways to work the algebra to solve these two equations.
One way is illustrated below:
Solve FOC (2) for X:
100 - 5X = 0
100 = 5X
X* = 100/5 = 20 (Note: we typically use a star "*" superscript to denote a solution value.)
Substitute X* back into FOC (1) and solve for :
2(1/(3X))3 - 5 = 0
2(1/(320))3 - 5 = 0
1/10 = 5
* = 1/50 = 0.02
Hence, the solution is: X* = 20 , * = 0.02.
(Note: Recall that in Lagrangian problems, the solution value of the Lagrangian multiplier, * = 0.02, is the
shadow value of the constraint.)
1
UNC-Wilmington ECN 321
Department of Economics and Finance Dr. Chris Dumas
Example (2):
* Nonlinear programming problem
* Two choice variables
* No interaction among choice variables in the utility function.
* One constraint
Note: When we have a nonlinear problem with one or more constraints, we need to use Lagrange's method)
max U = 2X1/2 + 3Y1/2
X,Y
subject to: 2X + 3Y 10000
Because this is a non-linear programming problem with a constraint, we will use Lagrange's method as the
solution method. Converting the problem into an equivalent Lagrangian problem, we re-write the objective
function as a Lagrangian expression, denoted "L." In doing so, we incorporate the constraint into the
Lagrangian expression by introducing the new choice variable into the problem as a Lagrangian
multiplier:
max L = 2X1/2 + 3Y1/2 + (10000 - 2X - 3Y)
X,Y,
subject to: nothing
F.O.C.'s:
L
(1) (1 / 2)2X 1 / 2 2 0
X
L
(2) (1 / 2)3Y 1 / 2 3 0
Y
L
(3) 10000 2X 3Y 0
The FOC equations provide us with three equations in three unknowns, X, Y and . Solve these three
equations for the solution values of X, Y and . There are several ways to work the algebra to solve these
three equations. One way is illustrated below:
Combine FOC's (1) and (2) to eliminate :
(1 / 2)2X 1 / 2 2
1 / 2
(1 / 2)3Y 3
2Y 1 / 2 2
3X 1 / 2 3
1/ 2
Y
1
X1 / 2
Y1 / 2 X1 / 2
Y 1/ 2 2 /1
X1 / 2 2 /1
Y = X call this equation (4)
2
UNC-Wilmington ECN 321
Department of Economics and Finance Dr. Chris Dumas
Substitute equation (4) back into FOC (3) and solve for X:
10000 - 2X - 3Y = 0
10000 - 2X - 3X = 0
10000 = 5X
X* = 2000
Substitute X* back into FOC (3) to find Y*:
10000 - 2X - 3Y = 0
10000 - 2*2000 - 3Y = 0
6000 = 3Y
Y* = 2000
Substitute X* back into FOC (1), or Y* back into FOC (2) , to find * :
(1 / 2) 2X 1 / 2 2 0
(1 / 2)2( 2000) 1 / 2 2
0.0224 2
* = 0.0112 (Note: The Lagrangian multiplier is the shadow value of the constraint.)
Example (3)
* Nonlinear programming problem
* Two choice variables
* One constraint
* Interaction among the choice variables in the utility function
Note: When we have a nonlinear problem with one or more constraints, we need to use Lagrange's method)
max U = (1+S)(1+P)
S,P
subject to: 25S + 8P 2000
Because this is a non-linear programming problem with a constraint, we will use Lagrange's method as the
solution method. Converting the problem into an equivalent Lagrangian problem, we re-write the objective
function as a Lagrangian expression, denoted "L." In doing so, we incorporate the constraint into the
Lagrangian expression by introducing the new choice variable into the problem as a Lagrangian
multiplier:
max L = (1+S)(1+P)+ (2000 - 25S - 8P)
S,P,
subject to: nothing
F.O.C.'s:
L
(1) 1 P 25 0
S
L
(2) 1 S 8 0
P
3
UNC-Wilmington ECN 321
Department of Economics and Finance Dr. Chris Dumas
L
(3) 2000 25S 8P 0
The FOC equations provide us with three equations in three unknowns, S, P and . Solve these three
equations for the solution values of S, P and . There are several ways to work the algebra to solve these
three equations. One way is illustrated below:
Combine FOC's (1) and (2) to eliminate :
1 P 25
1 S 8
25(1 S)
1 P
8
25(1 S)
P 1 call this equation (4)
8
Substitute equation (4) back into FOC (3) and solve for S:
2000 - 25S - 8P = 0
25(1 S)
2000 25S 8 1 0
8
2000 25S 25(1 S) 8 0
2000 25S 25 25S 8 0
1983 - 50S = 0
1983 = 50S
S* = 39.66
Substitute S* back into FOC (3) to find Y*:
2000 - 25S - 8P = 0
2000 - 25(39.66) - 8P = 0
1008.5 = 8P
P* = 126.06
Substitute S* and P* back into FOC (1) or (2) to find * :
1 + P - 25
=0
1 + 126.06 = 25
127.06 = 25
* = 5.08 (Note: The Lagrangian multiplier is the shadow value of the constraint.)
4
UNC-Wilmington ECN 321
Department of Economics and Finance Dr. Chris Dumas
Example (4)
* Nonlinear programming problem
* Two choice variables
* One constraint
* Interaction among the choice variables
Note: When we have a nonlinear problem with one or more constraints, we need to use Lagrange's method)
max U = 40X1/3Y1/4
X,Y
subject to: 2X + 3Y 2000
Because this is a non-linear programming problem with a constraint, we will use Lagrange's method as the
solution method. Converting the problem into an equivalent Lagrangian problem, we re-write the objective
function as a Lagrangian expression, denoted "L." In doing so, we incorporate the constraint into the
Lagrangian expression by introducing the new choice variable into the problem as a Lagrangian
multiplier:
max L = 40X1/3Y1/4+ (2000 - 2X - 3Y)
X,Y,
subject to: nothing
F.O.C.'s:
L
(1) (1 / 3)40X 2 / 3 Y1 / 4 2 0
X
L
(2) (1 / 4)40X1 / 3 Y 3 / 4 3 0
Y
L
(3) 2000 2X 3Y 0
The FOC equations provide us with three equations in three unknowns, X, Y and . Solve these three
equations for the solution values of X, Y and . There are several ways to work the algebra to solve these
three equations. One way is illustrated below:
Combine FOC's (1) and (2) to eliminate :
(1 / 3)40X 2 / 3 Y1 / 4 2
1/ 3 3 / 4 3
(1 / 4) 40X Y
4Y 2
3X 3
Y (1 / 2)X call this equation (4)
Substitute equation (4) back into FOC (3) and solve for X:
2000 2X 3Y = 0
2000 2X 3[(1/2)X] = 0
5
UNC-Wilmington ECN 321
Department of Economics and Finance Dr. Chris Dumas
2000 3.5X = 0
3.5X = 2000
X* = 2000/3.5 = 571.43
Substitute X* back into FOC (3) to find Y*:
2000 - 2X - 3Y = 0
2000 - 2(571.43) - 3Y = 0
857.14 - 3Y = 0
Y* = 857.14/3 = 285.71
Substitute X* and Y* back into FOC (1) or (2) to find * :
(1 / 3) 40X 2 / 3 Y1 / 4 2 0
(1 / 3)40X 2 / 3 Y1 / 4 2
2 / 3 1 / 4
(1 / 3) 40X Y
2
(1 / 3)40(571.43) 2 / 3 (285.71)1 / 4
2
* = 0.3980 (Note: The Lagrangian multiplier is the shadow value of the constraint.)
