Scribd
Upload
37 views

Dynamic Programming Notes

This document summarizes key concepts from Economics 101A Lecture 4: 1. It discusses constrained maximization problems and how to use the Lagrange multiplier method and implicit function theorem to solve them. Constrained optimization is important when there are limits imposed on choices. 2. The envelope theorem is introduced for how to calculate how a maximum changes with parameters in constrained optimization problems. 3. Preferences are discussed as the starting point for consumer theory, involving how individuals choose between different bundles of goods and activities based on their preferences.

Uploaded by

Shafia Akter Ranu
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
37 views

Dynamic Programming Notes

This document summarizes key concepts from Economics 101A Lecture 4: 1. It discusses constrained maximization problems and how to use the Lagrange multiplier method and implicit function theorem to solve them. Constrained optimization is important when there are limits imposed on choices. 2. The envelope theorem is introduced for how to calculate how a maximum changes with parameters in constrained optimization problems. 3. Preferences are discussed as the starting point for consumer theory, involving how individuals choose between different bundles of goods and activities based on their preferences.

Uploaded by

Shafia Akter Ranu
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 16

Economics 101A

(Lecture 4)

Stefano DellaVigna

January 29, 2009

Outline

1. Convexity and concavity II

2. Constrained Maximization

3. Envelope Theorem II

4. Preferences

Convexity and concavity


Alternative characterization of convexity.
A function f, twice dierentiable, is concave if and
only if for all x the subdeterminants |Hi| of the
Hessian matrix have the property |H1| 0, |H2|
0, |H3| 0, and so on.
For the univariate case, this reduces to f 00 0 for
all x
00 0 and
For the bivariate case, this reduces to fx,x

2
00
00
00
fx,x fy,y fx,y 0

A twice-dierentiable function is strictly concave if


the same property holds with strict inequalities.

Examples.
1. For which values of a, b, and c is f (x) = ax3 +
bx2 + cx + d is the function concave over R?
Strictly concave? Convex?
2. Is f (x, y) = x2 y 2 concave?
For Example 2, compute the Hessian matrix
fx0 =

, fy0 =

00 =
fx,x

00 =
, fx,y

00 =
fy,x

00 =
, fy,y

Hessian matrix H :
H=

00 =
fx,x
00
fy,x =

00 =
fx,y
00
fy,y =

00 and |H | = f 00 f 00
Compute |H1| = fx,x
2
x,x
y,y

2
00
fx,y

Why are convexity and concavity important?


Theorem. Consider a twice-dierentiable concave
(convex) function over C Rn. If the point x0
satisfies the fist order conditions, it is a global maximum (minimum).

For the proof, we need to check that the secondorder conditions are satisfied.

These conditions are satisfied by definition of concavity!

(We have only proved that it is a local maximum)

Constrained Maximization
Ch. 2, pp. 36-42 (3844, 9th Ed)
So far unconstrained maximization on R (or open
subsets)
What if there are constraints to be satisfied?
Example 1: maxx,y x y subject to 3x + y = 5
Substitute it in: maxx,y x (5 3x)
Solution: x =
Example 2: maxx,y xy subject to x exp(y)+y exp(x) =
5
Solution: ?

Graphical intuition on general solution.


Example 3: maxx,y f (x, y) = x y s.t. h(x, y) =
x2 + y 2 1 = 0
Draw 0 = h(x, y) = x2 + y 2 1.
Draw x y = K with K > 0. Vary K
Where is optimum?

Where dy/dx along curve xy = K equals dy/dx


along curve x2 + y 2 1 = 0
Write down these slopes.

Idea: Use implicit function theorem.

Heuristic solution of system

max f (x, y)
x,y

s.t. h(x, y) = 0

Assume:
continuity and dierentiability of h
h0y 6= 0 (or h0x 6= 0)

Implicit function Theorem: Express y as a function


of x (or x as function of y)!

Write system as maxx f (x, g(x))


g(x)

f.o.c.: fx0 (x, g(x)) + fy0 (x, g(x)) x = 0


g(x)

What is x ?

Substitute in and get: fx0 (x, g(x)) + fy0 (x, g(x))


(h0x/h0y ) = 0 or
h0x(x, g(x))
fx0 (x, g(x))
= 0
0
fy (x, g(x))
hy (x, g(x))

Lagrange Multiplier Theorem, necessary condition. Consider a problem of the type


max f (x1, x2, ..., xn; p)
x1,...,xn

h1 (x1, x2, ..., xn; p) = 0

h (x , x , ..., x ; p) = 0
n
2 1 2
s.t.

...

hm (x1, x2, ..., xn; p) = 0


with n > m. Let x = x(p) be a local solution to
this problem.

Assume:
f and h dierentiable at x
the following Jacobian matrix at x has maximal
rank

h1
(x ) ...
x1
J =
...
...

hm (x) ...
x1

h1
xn (x )

...

hm (x)
xn

Then, there exists a vector = (1, ..., m) such


that (x, ) maximize the Lagrangean function
L(x, ) = f (x; p)

m
X

j hj (x; p)

j=0

Case n = 2, m = 1.
First order conditions are

h(x; p)
f (x; p)

=0
xi
xi

for i = 1, 2

Rewrite as

fx0 1
fx0 2

h0x1

= 0
hx2

Constrained Maximization, Sucient condition


for the case n = 2, m = 1.
If x satisfies the Lagrangean condition, and the determinant of the bordered Hessian

h (x)
x
1
2
L (x)
2x1
2L (x)
x1x2

h (x)
x
2
2
L (x)
x2x1
2L (x)
x2x2

h (x)

H=
x1

h (x)
x
2
is positive, then x is a constrained maximum.

If it is negative, then x is a constrained minimum.


Why? This is just the Hessian of the Lagrangean L
with respect to , x1, and x2

Example 4: maxx,y x2 xy+y 2 s.t. x2 +y 2 p = 0

maxx,y, x2 xy + y 2 (x2 + y 2 p)
F.o.c. with respect to x:
F.o.c. with respect to y:
F.o.c. with respect to :
Candidates to solution?
Maxima and minima?

Envelope Theorem II
Envelope Theorem II: Ch. 2, pp. 42-43 (44, 9th Ed)
Envelope Theorem for Constrained Maximization. In problem above consider F (p) f (x(p); p).
We are interested in dF (p)/dp. We can neglect indirect eects:
m
hj (x(p); p)
f (x(p); p) X
dF
=

j
dpi
pi
pi
j=0

Example 4 (continued). maxx,y x2 xy + y 2 s.t.


x2 + y 2 p = 0
df (x(p), y (p))/dp?
Envelope Theorem.

Preferences
Part 1 of our journey in microeconomics: Consumer
Theory

Choice of consumption bundle:


1. Consumption today or tomorrow
2. work, study, and leisure
3. choice of government policy

Starting point: preferences.


1. 1 egg today 1 chicken tomorrow
2. 1 hour doing problem set 1 hour in class
... 1 hour out with friends
3. War on Iraq Sanctions on Iraq

Next Class
Properties of Preferences
From Preferences to Utility
Common Utility Functions

You might also like