0% found this document useful (0 votes)
48 views53 pages

Annotated Slides

Uploaded by

ge-guarin
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
48 views53 pages

Annotated Slides

Uploaded by

ge-guarin
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 53

Math Camp 2022

Redina Tahaj

Northeastern University

August 21, 2022

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 1 / 52
Derivatives

df (x) f (x + h) f (x)
= lim (1)
dx h!0 h
Derivative rules
Higher order derivatives

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 2 / 52
Extreme values

1 FOC: A point x ⇤ in the closed interval [a,b], such that f 0 (x)|x⇤ = 0 is


a local extreme value.
2 SOC:
I f 00 (x)|x ⇤ > 0 then x ⇤ is a local minimum.
I f 00 (x)|x ⇤ < 0 then x ⇤ is a local maximum.
Example: f (x) = x 3 + 12x 2 + 36x + 8

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 3 / 52
Concave and convex functions
Definition
A function f is concave i↵ x, x 0 2 X we have that
f (ax + (1 a)x) af (x) + (1 a)f (x 0) for all a 2 [0, 1]

The function is strictly concave if the above inequality is strictly satisfied.


Definition
A function f is convex i↵ x, x 0 2 X we have that
f (ax + (1 a)x)  af (x) + (1 a)f (x 0) for all a 2 [0, 1]

The function is strictly convex if the above inequality is strictly satisfied.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 4 / 52
Task: Prove that sum of two concave functions is a concave function

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 5 / 52
Concave Functions and the First Derivative
Theorem: A function is concave i↵:

f 0 (x 0)(x x 0) + f (x 0)  f (x) (2)

Let’s prove this preposition:

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 6 / 52
Concave Functions and Globality
Theorem:
If the function f (x) is concave, then an extreme value in the domain of
the function is a global maximum.

Proof: Using the condition for concavity evaluated at x 0,

f (x)  f (x 0) + f 0 (x 0)(x x 0) (3)

Because x 0 is an extreme value, we know that f 0 (x 0) = 0. Then we have:

f (x)  f (x 0) (4)

Remember that the definition of concavity is for every fair of points in the
domain of the function: thus it has to be that x 0 is greater than or equal
to the value of the function at any other point
Similarly we say that if the function is convex and it has an extreme value,
then that extreme value is a global minimum.
Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 7 / 52
Uniqueness

Strict Concavity and Uniqueness


Theorem: If x 0 is the global maximum of a strictly concave function then
x 0 is the unique global maximum.

Proof: By contradiction:
Suppose f has two maximizers, x and y, then tx + (1 t)y 2 G , and by the
definition of strict concavity, for 0 < t < 1 we have that:
f (tx + (1 t)y ) > tf (x) + (1 t)f (y ) = f (x) = f (y )
This contradicts the assumption that both x and y are global maximizers.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 8 / 52
Quasi-concave functions

Quasi-Concave Function
If f is quasi-concave then f (ax + (1 a)y ) min[f (x), f (y )]

Examples: f (x) = x , g (x, y ) = xy etc.


Quasi-Convex Function
If f is quasi-convex then f (ax + (1 a)y )  max [f (x), f (y )]

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 9 / 52
Quasi-concavity and the first derivative
Theorem: A function f is quasi-concave i↵:

f 0 (x 0)(x x 0) > 0 (5)

Let’s prove it:

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 10 / 52
Globality, uniqueness, and quasi-concave/quasi-convex
functions

1 Strict quasi-concavity and global maximum: If f (x) is strictly


quasi-concave and x 0 is an extreme value in the domain of the
function, then x 0 is a global maximum.
2 Strict quasi-concavity and uniqueness: If x 0 is the global
maximum of a strictly quasi-concave function, we know that x 0 is the
unique maximum.
3 Likewise, for strictly quasi-convex functions, if there exists an extreme
value it is the unique global minimum.
4 Globality and uniqueness are not determined for extreme values of
quasi-concave or quasi-convex functions.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 11 / 52
Practice:
p
Determine if the function f (x) = x is concave, convex, or not.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 12 / 52
Summary
Function Globality Uniqueness
Concave
Convex
Strictly Concave
Strictly Convex
Quasi-concave
Quasi-convex
Strictly quasi-concave
Strictly quasi-convex

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 13 / 52
Gradients and Hessians
Gradient is the vector of first derivatives.

© x 1™
f
≠ fx 2 Æ
5f (x) = ≠≠ . ÆÆ
≠ .. Æ̈
´fx n
Hessian is the matrix of second derivatives.
© x1.x1 ™
f . . . fx1 xk

H (f (x)) = ≠ .. Æ
Æ̈
´fxk x1 . . . fxk xk
Young’s Theorem fxy = fyx

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 14 / 52
Finding Extreme Values of Functions with n variables
FOC: A vector x ⇤ in the closed set A, such that 5f (x) = 0 is a local
extreme value.
SOC: A vector x ⇤ in the closet set A, such that H (x ⇤ ) is negative
semidefinite is a local maximum, while a vector x ⇤ in the closet set A, such
that H (x ⇤ ) is positive semidefinite is a local minimum. Further if f (x) is
concave or strictly quasi-concave, then the extreme value is a global
maximum and if f (x) is convex or strictly quasi-convex, the extreme value
is a global minimum. As in the case of functions with one variable,
uniqueness is guaranteed by strictly concave, strictly convex, strictly
quasi-concave or strictly quasi-convex functions.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 15 / 52
Constrained Optimization with equality constraints
Equality constraint
An equality constraint is a condition that must be satisfied in a constrained
optimization problem such that: g (x) = c where c is a constraint.

Maximization with equality constraints


Given a set of M equality constraints, an objective function f (x) with
domain R N . The problem of maximization with equality constraints is
defined as: Max f (x) subject to, g1 (x) = c1 ... gM (x) = cM

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 16 / 52
Lagrange Method

Lagrangian:
Given the maximization problem with equality constraints, the
corresponding Lagrangian L(x, _) is defined as:
ÕM
L(x, _) = f (x) + i=1 _i (ci gi (x))

_ is a set of M variables called Lagrange multipliers.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 17 / 52
Lagrangian Sufficiency Theorem:
If x ⇤ and _⇤ are local maximizers of the Lagrangian then x ⇤ has satisfied
the constraints and locally maximized f (x).
FOC: 5L = 0
To make sure that x ⇤ is a local maximum we need to use a SOC: If the
Bordered Hessian is negative semidefinite we have a local maximum.
Globality: f (x) has to be at least concave or strictly quasi-concave and
the upper contour set of g (x) be compact and convex.
Uniqueness: f (x) be at least strictly concave or strictly quasi-concave
and the upper contour set of g (x) be compact and convex

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 18 / 52
Steps:
Consider optimizing f (x, y ) s.t. g (x, y ) = c.
1 Define L(x, y , _) = f (x, y ) + _(g (x, y ) c)
2 FOC: 5L = 0
© ™
Lxx Lxy Lx_
3 Form the bordered-Hessian: BH= ≠ Lyx Lyy Ly _ Æ̈
´L_x L_y L__
4 Evaluate BH at critical points and find its determinant
5 If the determinant is postive then we have a local maximum, if
negative then we have a local minimum.
6 To determine whether the optimal solution is unique or global you
would have to check the features of function f (x, y )

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 19 / 52
Example:
p
Maximize xy s.t. x + y = 2

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 20 / 52
Constrained Optimization with Inequality Constraints
Given a set of M inequality constraints, an objective function f (x) with
domain RN , the problem of maximization with inequality constraints is
defined:
Max f (x) s.t. gM (x)  cM
Solve using a generalization of the Lagrange method
The Karush-Khun-Tucker (KKT) Method:
If x ⇤ and _⇤ are local maximizers of the Lagrangian and satisfy the
conditions:
1 _⇤ (c g (x)) 0
2 _⇤ 0
Then x ⇤ is a local maximizer

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 21 / 52
KKT

With this method we need two conditions to be satisfied:


Õm
1 FOC: 5f (x) =
i=1 _ i 5 gi (x)
2 Slackness conditions:

I If _ i 0 and _i (c g (x)) = 0 we have a maximum.


I If _ i  0 and _i (c g (x)) = 0 we have a minimum.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 22 / 52
Let’s look into an example:
A consumer lives on an island where she produces two goods, x and y,
according to her production possibility frontier x 2 + y 2  200, and she
consumes all goods herself. Her utility function is xy 3 . The consumer also
faces environmental constraint on her total output of both goods. The
environmental constraint is given by x + y  20.
1 Write the KKT first order conditions
2 Find the optimal solution. Determine which restrictions are binding.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 23 / 52
’Cont.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 24 / 52
The Envelope Theorem

Theorem: Suppose that the constraint optimization problem

max F (x, y , b) s.t. g (x, y , b) = c (6)

admits a unique solution (x ⇤ (b), y ⇤ (b), b) with an associated Lagrange


multiplier _⇤ (x ⇤ , y ⇤ , b). If L(x, y , _, b) denotes the Lagrangian then,

dF (x ⇤ (b), y ⇤ (b), b)/d b = mL/mb | (x ⇤ ( b ),y ⇤ ( b ), b ) (7)

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 25 / 52
Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 25 / 52
Applications of the Envelope Theorem

It is used in policy analysis in general, e/g e↵ects of wealth redistribution


on welfare.
Microeconomic theory: Hicksian demand and Expenditure Function (MWG
p.69), Roy’s Identity (MWG p.74), Shephard’s Lemma (MWG p.141),
Hotelling’s Lemma (MWG p.138)

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 26 / 52
Implicit Function Theorem

Let F (x; U) be a function in a neighborhood of (x ⇤ , U⇤ ) in R 2 . Suppose


that F (x ⇤ ; U⇤ ) = 0 and consider the expression:

F (x, U) = 0 (8)

If mF (x ⇤ , U⇤ )/mx < 0, then there exists a function x = x (U) defined on an


open interval I about a point U = U⇤ such that:
1 F (x (U), U) = 0 for all U 2 I
2 x (U⇤ ) = x
3 dx
(U ⇤) = mF
[x (U ⇤ ); U ⇤ ]/ mF [x (U⇤ ); U]
dU mU mx

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 27 / 52
Example

Let us consider a firm that produces a good y by using a single input x.


The firm sells the output and acquires the input in competitive markets:
The market price of y is p, and the cost of each unit of x is just w. Its
technology is given by f : R+ ! R+; where f (x) = x a and a 2 (0; 1). Its
profits are given by
c(x, p, w ) = px a wx (9)
The firm selects the input level, x; in order to maximize profits. We would
like to know how its choice of x is a↵ected by a change in w.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 28 / 52
Assuming an interior solution, the FOC of this optimization problem is:

Notice that F (x, p, w ) = pa(x ⇤ ) a 1 w


mF (x ⇤ ,p,w )
Since mx = pa(a 1) (x ⇤ ) a 1 < 0 we can use the above theorem and
get:
dx
(p, w ) = (10)
dw

Thus if the price of the input increases, then the firm will acquire less of it.
That is the unconditional input demand is decreasing in the input price.
See examples and further discussion in Dropbox folder notes.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 29 / 52
Review of Integration Rules
π
d
F (x) = f (x) and f (x)dx = F (x) + c (11)
dx
Basic Rules of Integration
Power Rule

Exponential Rule

Logarithmic Rule
1 π
0
f (x)e f (x) dx = e f (x) + c (12)
2
π 0
f (x)
d (x) = ln | f (x) | +c (13)
f (x)
Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 30 / 52
’Cont.
Integral of a sum

Integral of a multiple

Substitution Rule

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 31 / 52
Integration by parts
π π
vdu = uv udv (14)
Ø Ø
Example: x (x + 1) 1/2 dx and lnxdx

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 32 / 52
Definite Integrals
π b
f (x)dx = F (x) | ba = F (b) F (a) (15)
a
Properties of Definite Integrals
Øa Øb
1
b
f (x)dx = a f (x)dx

Øa
2
a
f (x)dx = 0

Ød Øb Øc Ød
3
a
f (x)dx = a
f (x)dx + b
f (x)dx + c
f (x)dx

4 Besides these, the same rules that apply to indefinite integrals also
apply to definite integrals.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 33 / 52
Improper Integrals
π 1 π b
f (x)dx or f (x)dx (16)
o 1
π 1 π b
f (x)dx = lim f (x)dx (17)
a b!1 a
If this limit exists integral is said to be convergent, while if the limit does
not exist it is said to be divergent (meaningless).
Infinite Integrand: Even with finite bounds an integral can still be
improper is the integrand becomes infinite somewhere in [a,b].
Ø11
e/g. 0 x dx As x ! 0 x1 ! 1
Ø1 1
To solve we still use limits. a x
dx = ln | 1a = ln1 lna = lna (diverg.)
Economic Applications of using integrals involve finding total functions when
marginal ones are given, finding the present value of cash flows etc.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 34 / 52
First-Order Di↵erential Equations
Contains only the first derivative (but it may have various powers). The
highest power is called the degree.
1 Constant coefficient and constant term.
dy
+ u(t)y = w (t) (18)
dt

dy
I Homogeous case dt + ay = 0

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 35 / 52
First-Order Di↵erential Equations
Non-homogenous case dydt + ay = b Solution consists of the
complementary function (solution to the reduced function) and the
particular integral.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 36 / 52
Example
dy
Solve dt + 2y = 6

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 37 / 52
Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 38 / 52
An Economic Application
Dynamics of Market Price:

Qd = U VP and Qs = W + XP (19)
If P (0) < P ⇤ does the time path of P (t) converge to P ⇤ as t ! 1 ?
dP
Assume dt = j (Qd Qs )

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 39 / 52
An Economic Application
P (t) = [P (0) P ⇤ ]e kt + P⇤
k > 0 P (t) ! 0 as t ! 1

Dynamic Stability requires the asymptotic vanishing of the complementary


function as t becomes infinite. So to assure dynamic stability in this
setting we must have that k < 0 or j (V + X) > 0.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 40 / 52
Di↵erence Equations
Discrete time means that the value of y will change only when
variable t changes from one integer value to the next, referring to as
periods. The discrete-time version of economic dynamics is often
referred to as period analysis.
The pattern of change of the variable y over time will be represented
by the di↵erence quotient yt which is the discrete-time counterpart
of dy
dt . Since t = 1 , can be simplified to y = yt+1 yt called the
first di↵erence of y .
Iterative Method of solving a first order di↵erence equation:
Example: yt+1 = yt + 2 and y0 = 15

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 41 / 52
General Method: yt+1 + ayt = c where a and c are constants. The
solution involves a particular solution (any solution to our equation)
and the complementary solution (solution to the reduced form). Here
yp represents again the inter-temporal equilibrium level of y while yc
represents the deviation of the time path from that equilibrium.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 42 / 52
General Method
To find the solution to the complementary function here too we try a
solution of the form yc = Ab t
Thus yt+1 = Ab t+1

For yp we need any solution to the given equation. We try the simplest
form, a constant. yt = k Thus yt+1 = k too.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 43 / 52
General Method
Now if a = 1 we must try another kind of solution such as yt = kt. Let’s
see how this works.

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 44 / 52
Example
Solve yt+1 5yt = 1 y0 = 7/4

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 45 / 52
Dynamic Stability of the Equilibrium

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 46 / 52
Dynamic Stability of the Equilibrium

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 47 / 52
Dynamic Stability of the Equilibrium

Whether the equilibrium is dynamically stable is whether or not Ab t will


tend to vanish as t ! 1.

The time path of b t will be:


1 Non oscillatory if b > 0
2 Oscillatory if b < 0
3 Divergent if | b |> 1
4 Convergent if | b |< 1

The constant A produces a scale e↵ect without changing the basic


configuration of the time path. If A is negative the time path will be a
mirror image.

In sum, the solution yt = Ab t + yp will be convergent i↵ | b |< 1

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 48 / 52
Example
4 t
Example: Time path described by yt = 2( 5) +9

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 49 / 52
An Economic Application: the Cobweb Model
Consider a model where the quantity supplied is a function if the price in
the preceding time period.
Take Qst = W + XPt 1 while Qdt = U VPt
Market clearing: Qst = Qdt
Shift by one period: VPt+1 + XPt = U + W

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 50 / 52
The Cobweb Model
Lets see the graphs of the time path of this model in two cases, when the
supply curve is steeper than the demand curve (explosive oscillation) and
vice-versa (convergence to inter-temporal price).

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 51 / 52
Resources

If interested, please find further readings on the Dropbox folder under


Chiang’s (Dynamic Optimization).

Redina Tahaj (Northeastern University) Math Camp 2022 August 21, 2022 52 / 52

You might also like