GARP and Afriats Theorem

Production

Econ 2100 Fall 2017

Lecture 8, September 21

Outline
1 Generalized Axiom of Revealed Preferences
2 Afriats Theorem
3 Production Sets and Production Functions
4 Prots Maximization, Supply Correspondence, and Prot Function
5 Hotelling and Shephard Lemmas
6 Cost Minimization
 From Last Class

fx j ; p j ; w j gNj=1 is a nite set of demand data that satisfy p j x j wj.

Directly revealed weak preference: x j %R x k if pj x k wj
Directly revealed strict preference: x j R
xk if p j x k < w j
x j %R x i 1 ;
if there exists x i 1 %R x i 2 ;
Indirectly revealed weak preference:
x j %I x k x i1 ; x i2 ; : : : ; x im :::
such that x i m %R x k
Indirectly revealed strict preference:
x j %R x i1 ;
if there exists x i1 %R x i2 ; and one of the relations
j I k
x x x i1 ; x i2 ; : : : ; x im :::
such that x im %R x k in the chain is strict
Using these denitions, we can nd a conditions that guarantees choices are
the result of the maximizing a preference relation or a utility function.
 Generalized Axiom of Revealed Preference
Axiom (Generalized Axiom of Revealed Preference - GARP)
If x j %R x k , then not x k I
xj .

If a choice is directly revealed weakly prefered to a bundle, this bundle cannot

be indirectly revealed strictly preferred to that choice.
Note that if p j x j < w j then x j R x j and therefore x j I x j and therefore
GARP cannot hold.
Therefore, GARP implies the consumer spends all her money.
Now suppose x j %R x k and x k I x j . Then there is a chain of weak preference
from j to k, and a chain back (from k to j) that has at least one strict
preference.
This is a cycle with a strict preference inside; if such a cycle exists one can
construct a money pump to make money o the consumer and leave her
exactly at the bundle she started from.
GARP rules these cycles out.

Problem 4, Problem Set 4.

Show that GARP is equivalent to the following: If x j %I x k then not x k I
xj .
 Utility and Demand Data

Denition
A utility function u : Rn+ ! R rationalizes a nite set of demand data
fx j ; p j ; w j gNj=1 if
u(x j ) u(x) whenever pj x wj

This denition can alternatively be stated by the condition

x j 2 arg maxn u(x) subject to p j x wj
x 2R+

If the observations satisfy the denition, the data is consistent with the
behavior of an individual who, in each observation, chooses a
utility-maximizing bundle subject to the corresponding budget constraint.
 Afriats Theorem
Theorem (Afriat)
Given a nite set of demand data f(x j ; p j ; w j )gNk=1 , the following are equivalent:
1 There exists a locally nonsatiated utility function which rationalizes the data.
2 The data satisfy the Generalized Axiom of Revealed Preference.
3 There exist numbers v j ; j > 0 such that
j
vk vj + [p j x k w j ]; for all j; k = 1; : : : ; N (G)
4 There exists a continuous, strictly increasing, and concave utility function
which rationalizes the data.

For nite data, the only implication of preference maximization is GARP;

conversely, if the data violate GARP, we conclude the consumer is not
maximizing any continuous, strictly monotone, and convex preference relation.
Using a nite number of observations, one cannot distinguish a continuous,
strictly monotone, and convex preference from a locally nonsatiated one.
One way to test if demand data falsify utility maximization is to see if one can
nd the numbers in (G).
These numbers help us construct a function that rationalizes data: each
observation has utils v i and a budget violating penalty/prize i ; every
choice is rational given these.
 Afriats Theorem

Theorem (Afriat)
Given a nite set of demand data f(x j ; p j ; w j )gNk=1 , the following are equivalent:
1 There exists a locally nonsatiated utility function which rationalizes the data.
2 The data satisfy the Generalized Axiom of Revealed Preference.
j
3 There exist numbers v j ; > 0 such that
k j j
v v + [p j x k w j ]; for all j; k = 1; : : : ; N (G)
4 There exists a continuous, strictly increasing, and concave utility function
which rationalizes the data.

Proof Strategy
The fact (4) implies (1) is immediate (strictly increasing implies locally non
satiated).
We will skip the remainder of the proof, but see Kreps or Varian for it.
 The Afriat Numbers
j j
9v j ; > 0 s.t. v k vj + [p j x k w j ] for all j; k = 1; : : : ; N (G)

k
One uses the numbers v k and to construct a function that rationalizes the
data;
the theorem then says that every choice is rational.

Where are the Afriats numbers coming from?

Consider utility maximization subject to a budget constraint:
x 2 arg maxn u(x) s.t. p x w:
x 2R+

The solution maximizes the Lagrangean:

L(x; ) = u(x) + (w p x):
Let x and be the optimal consumption and multiplier values, respectively.
Then we have L(x ; ) L(x ; ) for all x 2 Rn+ . That is,
u(x ) + (w p x ) u(x ) + (w p x ):
| {z }
=0 since budget constraint binds
Thus for all x 2 Rn+
u(x ) u(x ) + (p x w)
 Intuition Behind the Afriat Numbers
j j
9v j ; > 0 s.t. v k vj + [p j x k w j ] for all j; k = 1; : : : ; N (G)

Another intuition behind Afriats numbers

Rewrite (G) as
j j
vk vj [p x k w j ]:
Suppose that p j x k w j > 0. Then x k was unaordable at p j and w j .
Even if x k yields more utils than x j , at p j and w j the penalty for violating the
budget constraint (given by j [p j x k w j ]) is too large for the change in the
value of the choice (given by v k v j ) to compensate for it.
Thus, the decision maker chose x j over all other alternatives at p j and w j .
Suppose that p j x k wj 0. Then x k was aordable at p j and w j .
But then (G) implies that
vj vk j
[w j p j x k ]:
j k
So the decision maker chose x over x because the dierence in utils is
higher than the bonus from having some extra money left over by choosing x k
instead of x j .
 Afriats Theorem
Summary
For nite data sets, preference maximization is equivalent to GARP.
With GARP, we can nd utility values for every data point (Afriat numbers).
The decision maker behaved as if each chosen data point was optimal (i.e.
utility-maximizing).
The theorem also gives us a constructive proof of the convexity of the upper
contour sets of any element in the data set.
These are convex since they are the intersections of half spaces (this comes from
the proof that (3) implies (4)); however, the utility representation is concave.
In general convex upper contour sets are associated with convex preferences and
quasiconcave utility.
Therefore using nite demand data one cannot test convexity of preferences.

Using a nite number of observations we cannot distinguish a consumer with

continuous, strictly monotone, and convex preferences from a consumer with
locally nonsatiated preference (the extra assumptions have absolutely no bite
for nite data sets).
 Producers and Production Sets
Producers are prot maximizing rms that buy inputs and use them to produce and
then sell outputs.

The plural is important because most rms produce more than one good.

The standard textbook description in intermediate microeconomics focuses on

one ouput and a few inputs (two in most cases).
Usually, you may have seen production described by a function that takes
inputs as the domain and output(s) as the range.
Here we focus on a more general and abstract version of this same idea.

Denition
A production set is a subset Y Rn .

y = (y1 ; :::; yN ) denotes production (input-output) vectors.

A production vector has outputs as non-negative numbers and inputs as
non-positive numbers: yi 0 when i is an input, and yi 0 if i is an output.
What is p y in this notation?
 Production Set Properties

Denition
Y Rn satises:
no free lunch if Y \ Rn+ f0n g;
possibility of inaction if 0n 2 Y ;
free disposal if y 2 Y implies y 0 2 Y for all y 0 y;
irreversibility if y 2 Y and y 6= 0n imply y2
= Y;
nonincreasing returns to scale if y 2 Y implies y 2 Y for all 2 [0; 1];
nondecreasing returns to scale if y 2 Y implies y 2 Y for all 1;
constant returns to scale if y 2 Y implies y 2 Y for all 0;
0 0
additivity if y ; y 2 Y imply y + y 2 Y ;
convexity if Y is convex;
Y is a convex cone if for any y ; y 0 2 Y and ; 0, y + y 0 2 Y .

Draw Pictures.
 Production Set Properties Are Related

Some of these properties are related to each other.

Exercise
Y satises additivity and nonincreasing returns if and only if it is a convex cone.

Exercise
For any convex Y Rn such that 0n 2 Y , there is a convex Y 0 Rn+1 that
satises constant returns such that Y = fy 2 Rn : (y ; 1) 2 Rn+1 g.
 Production Functions
Let y 2 Rm n
+ denote outputs while x 2 R+ represent inputs; if the two are related by
n m
a function f : R+ ! R+ , we write y = f (x) to say that y units of outputs are
produced using x amount of the inputs.

When m = 1, this is the familiar one output many inputs production function.
Production sets and the familiar production function are related.

Exercise
Suppose the rms production set is generated by a production function
f : Rn+ ! Rm n
+ , where R+ represents its n inputs and R+ represents its m outputs.
Let
Y = f( x; y ) 2 Rn Rm
+ :y f (x)g:
Prove the following:
1 Y satises no free lunch, possibility of inaction, free disposal, and irreversibility.
2 Suppose m = 1. Y satises constant returns to scale if and only if f is
homogeneous of degree one, i.e. f ( x) = f (x) for all 0.
3 Suppose m = 1. Y satises convexity if and only if f is concave.
 Transformation Function
We can describe production sets using a function.
Denition
Given a production set Y Rn , the transformation function F : Y ! R is dened
by
Y = fy 2 Y : F (y ) 0 and F (y ) = 0 if and only if y is on the boundary of Y g ;
the transformation frontier is
fy 2 Rn : F (y ) = 0g

Denition
Given a dierentiable transformation function F and a point on its transformation
frontier y , the marginal rate of transformation for goods i and j is given by
@F (y )
@yi
MRTi ;j = @F (y )
@yj

Since F (y ) = 0 we have
@F (y ) @F (y )
dyi + dyj = 0
@yi @yj
MRT is the slope of the transformation frontier at y .
 Prot Maximization

Prot Maximizing Assumption

The rms objective is to choose a production vector on the trasformation frontier
as to maximize prots given prices p 2 Rn++ :
max p y
y 2Y
or equivalently
max p y subject to F (y ) 0

Does this distinguish between revenues and costs? How?

Using the single output production function:
max pf (x) w x
x 0

here p 2 R++ is the price of output and w 2 Rl++ is the price of inputs.
 First Order Conditions For Prot Maximization
max p y subject to F (y ) = 0

Prot Maximizing
The FOC are
@F (y )
pi = for each i or p = rF (y ) in matrix form
@yi |{z} | {z }
1 n 1 n
Therefore
@F (y )
1 @yi
= for each i
pi
the marginal product per dollar spent or received is equal across all goods.
Using this formula for i and j:
@F (y )
@yi pi
@F (y )
= MRTi ;j = for each i; j
pj
@yj

the Marginal Rate of Transformation equals the price ratio.

 Supply Correspondence and Prot Functions
Denition
Given a production set Y Rn , the supply correspondence y : Rn++ ! Rn is:
y (p) = arg max p y :
y 2Y

Tracks the optimal choice as prices change (similar to Walrasian demand).

Denition
Given a production set Y Rn , the prot function : Rn++ ! R is:
(p) = max p y :
y 2Y

Tracks maximized prots as prices change (similar to indirect utility function).

Proposition
If Y satises non decreasing returns to scale either (p) 0 or (p) = +1.

Proof.
Question 4, Problem Set 5.
 From Last Class
Denition
A production set is a subset Y Rn .

Denition
Given a production set Y Rn , the transformation function F : Y ! R is
Y = fy 2 Y : F (y ) 0 and F (y ) = 0 if and only if y is on the boundary of Y g ;
the transformation frontier is fy 2 Rn : F (y ) = 0g.

A rms objective is to choose an output vector in its production set so as to

maximize prots.
Denition
Given a production set Y Rn , the supply correspondence y : Rn++ ! Rn is:
y (p) = arg max p y :
y 2Y

Denition
Given a production set Y Rn , the prot function : Rn++ ! R is:
(p) = max p y :
y 2Y
 Properties of Supply and Prot Functions

Proposition
Suppose Y is closed and satises free disposal. Then:
( p) = (p) for all > 0;
is convex in p;
y ( p) = y (p) for all > 0;
if Y is convex, then y (p) is convex;
if jy (p)j = 1, then is dierentiable at p and r (p) = y (p) (Hotellings
Lemma).
if y (p) is dierentiable at p, then Dy (p) = D 2 (p) is symmetric and
positive semidenite with Dy (p)p = 0.
 The Prot Function Is Convex
Proof.
Let p,p 0 2 Rn++ and let the corresponding prot maximizing solutions be y and y 0 .
For any 2 (0; 1) let p = p + (1 ) p 0 and let y be the prot maximizing
output vector when prices are p.
By revealed preferences
p y p y and p0 y 0 p0 y
why?
multiply these inequalities by and 1
p y p y and (1 ) p0 y 0 (1 ) p0 y
summing up
p y + (1 ) p0 y 0 [ p + (1 ) p0] y
using the denition of prot function:
(p) + (1 ) (p 0 ) ( p + (1 ) p0)
proving convexity of (p).
 The Supply Correspondence Is Convex
Proof.
Let p 2 Rn++ and let y ; y 0 2 y (p).
We need to show that if Y is convex then
y + (1 ) y 0 2 y (p) for any 2 (0; 1)
By denition:
p y p y for any y 2 Y and p y0 p y for any y 2 Y

multiplying by and 1 we get

p y p y and (1 ) p y0 (1 )p y
Therefore, summing up, we have
p y + (1 ) p y0 [ + (1 )] p y
Rearranging:
p [ y + (1 ) y 0] p y
proving convexity of y (p).
 Hotellings Lemma
if jy (p)j = 1, then is dierentiable at p and r (p) = y (p)
Proof.
Suppose y (p) is the unique solution to max p y subject to F (y ) = 0.
The Envelope Theorem says
>
Dq (x (q); q) = Dq (x; q)jx =x (q);q=q [ (q)] Dq F (x; q)jx =x (q);q=q
In our setting,
(x ; q) = p y ,
F (x ; q) = F (y ), and
(x (q); q) = (p).
Thus, by the envelope theorem:
> >
r (p) = Dp (p y )jy =y (p) [ (p)] Dp F (y )jy =y (p) = y jy =y (p) [ (p)] 0

because p y is linear in p and Dp F (y ) = 0.

Therefore
r (p) = y (p)
as desired.
 Law of Supply
Remark
If y (p) is dierentiable at p, then Dy (p) = D 2 (p) is positive semidenite.

Write the Lagrangian

L=p y F (y )
By the Envelope Theorem:
@ (p) @L
= = yi (p)
@pi @pi y =y

Therefore, we have
@ 2 (p) @y (p)
= i 0
@pi @pi

where the inequality follows from convexity of the prot function.

This is called the Law of Supply: quantity responds in the same direction as
prices.
Notice that here yi can be either input or output.
What does this mean for outputs?
What does this mean for inputs?
 Factor Demand, Supply, and Prot Function
The previous concepts can be stated using the production function notation.

Denition
Given p 2 R++ and w 2 Rn++ and a production function f : Rn+ ! R+ , the rms
factor demand is
x (p; w ) = arg max fpy w x subject to f (x) = y g = arg max pf (x) w x:
x x

Denition
Given p 2 R++ and w 2 Rn++ and a production function f : Rn+ ! R+ , the rms
supply y : Rn+ ! R is dened by
y (p; w ) = f (x (p; w )) :

Denition
Given p 2 R++ and w 2 Rn++ and a production function f : Rn+ ! R+ , the rms
prot function : R++ Rn++ ! R is dened by
(p; w ) = py (p; w ) w x (p; w ) :
 Factor Demand Properties

Given these denitions, the following results translate the results for output
sets to production functions.

Proposition
Given p 2 R++ and w 2 Rn++ and a production function f : Rn+ ! R+ ,
1 (p; w ) is convex in (p; w ).
@y (p;w )
2 y (p; w ) is non decreasing in p (i.e. @p 0) and x (p; w ) is non
@xi (p;w )
increasing in w (i.e. @w i 0) (Hotellings Lemma).

Proof.
Problem 2a,b; Problem Set 5.
 Cost Minimization
Cost Minimizing
Consider the one output case and suppose the rm wants to deliver a given
output quantity at the lowest possible costs. The rm solves

min w x subject to f (x) = y

This has no simple equivalent in the output vector notation.

Denition
Given w 2 Rn++ and a production function f : Rn+ ! R+ , the rms conditional
factor demand is
x (w ; y ) = arg min fw x subject to f (x) = y g ;

Denition
Given w 2 Rn++ and a production function f : Rn+ ! R+ , the rms cost function
C : Rn++ R+ ! R is dened by
C (w ; y ) = w x (w ; y ) :
 Properties of Cost Functions
Proposition
Given a production function f : Rn+ ! R+ , the corresponding cost function C (w ; y )
is concave in w .

Proof.
Question 2c; Problem Set 5. (Hint: use a revealed preferencesargument)

Shephards Lemma
Write the Lagrangian
L=w x [f (x) y]
by the Envelope Theorem
@C (w ; y ) @L
= = xi (w ; y )
@wi @wi

Conditional factor demands are downward sloping

2
(w ;y ) @xi (w ;y )
Dierentiating one more time: @C @w i @w i = @w i 0 where the
inequality follows by Question 2c in Problem Set 5.