Lecture 3: Last updated 9/03/2010 CHAPTER 11 (part 1) SPECIAL TOPICS IN CONSUMER THEORY Revealed Preferences: “So far we have
approached demand theory by assuming the consumer has preferences satisfying certain properties (complete, transitive, and strictly monotonic); then we have tried to deduce all the observable properties of market demand that follow as a consequence (budget balances, symmetry, and negative semidefinitenes of the Slutsky matrix). Thus we have begun by assuming something about things we cannot observe, preferences, to make predictions about something we can observe, consumer demand behavior. In his remarkable Foundations of Economic Analysis, Paul Samuelson suggests an alternative approach. Why not start and finish with observable behavior? A consumer’s observable market behavior can also be derived from few simple and sensible assumptions about consumer’s observable choices, (rather than unobservable preferences). The basic idea is simple; if consumer buys one bundle instead of another affordable bundle, then the first bundle is considered to be revealed preferred to the second.” (Jehle, G.A., Reny, P.J., Advanced Microeconomic Theory, 2nd Edition, Addison Wesley, pp. 86-87) Then; Is it possible to replace the utility maximization hypothesis with one based entirely on observable quantities (consumed by the consumer)? OR Is it possible starting with a set of demand relations which obey the symmetry and negative-semidefiniteness of the pure substitution terms to infer that there exist some utility functions from which those demand functions are derivable? (Integrability) Example: there are two goods
y1 Y2 y(y1,y2)
... Then x1 will never be revealed preferred to x0.... Then it is said that x0 is revealed preferred to x1........ y n ) P = (P1 .........Pn )
The vector notation x2
∑ x P = xP = M
will be used
0 P10 x10 + P20 x2
0 1 0 1 0 0 0 0 P 1 x1 + P 2 x2 < P 1 x1 + P 2 x2
x1 Although x0 is more expensive than x1 (in prices of x0) and he can afford x1.. The axiom can be formulated as..... the consumer chooses x0.. P1 x1 < P1 x0 Contradicting Case This case is against the theory......P1 x1 + P2 x2 ≥ P1 y1 + P2 y2 ==> x is revealed to preferred to y
x = (x1 .... If P0x0 ≥ P0x1..... y 2 .....x n ) y = ( y1 ....... i.
then when x1 is chosen (at the price level P1)..... Weak Axiom of Revealed Preferences (WARP) Assume x0 is revealed preferred to x1.... x0 is chosen and P0x0 ≥ P0x1.... P2 .. at P0.......e.
........ x 2 ..
(Meaning of Pdx = 0? This is what is implied by “utility is held constant”. they are functions). .. Proof: Proof must be read from the book. . So dU = 0 . A square matrix aij is said to be negative semidefinite if
∑∑ h a h
≤ 0 for all hi. M ) are homogenous of degree zero in prices and income. (can you see this in the diagram above?) Now we will try to see which of the results (homogeneity of the demand functions.. Proposition 1: demand relations xi = x1M (P1 . Pn . Weak axiom also implies that Hicks – Slutsky substitutution terms are negative. hj. Pn . Proposition 2: weak axiom implies that the relations x1 = x1M (P1 . Proof must be read from the book....
.. implies dU = U1dx1 + U 2 dx 2 = λ (P 1dx1 + P 2 dx 2 )
∑ Pdx = 0 . negativity of the pure substitution terms. M ) are single valued (so. Silberberg pp 318.. Defn: Negative semidefinite matrix. Proposition 3: The matrix Sij is negative semidefinite under the weak axiom of revealed preferences. Silberberg pp 319..Note that x0 is revealed preferred to x1 and x1 is revealed preferred to x0. the single valuedness and homogeneity of degree zero of demand functions are implied by the weak axiom. Proof: Proof must be read from the book. Silberberg pp 320. since λ ≠ 0 . .. symmetry of the Slusky terms ) implied by utility maximization are also implied by the weak axiom of revealed preferences.
2) 1 ⎞ ⎛ ⎛ 1 ⎞ x 2 = ⎜ 4. you will see that (1)
P 0 x 0 = P 0 x1 = 12
so x 0 is revealed preferred to x1 . Hence dpdx ≤ 0 ==> own substitution effect is negative. P 2 x 2 < P 2 x1 ==> 17 < 17. the weak axiom does not imply the symmetry of the Slutsky terms.dx would be interpretable as “pure substitution effect” (so on the same indifference curve). So the inconsistent consumer.1 . when x 2 is chosen.
P 0 x 0 = 12 < P 0 x 2 = 13
x2 is revealed preferred to x0.5 P2x2=17
Then. x0 is revealed preferred to x2 and then when x2 is chosen But P 2 x 2 < P 2 x 0 we see that this is not the case.2) x1 = (3.3. Consider the choices of the consumer below.2) P 0 = (2.5 ==> x1 > x 2
we expect that x 0 > x 2 i. P 0 x 0 = P 0 x 2 = 17 .e.2.
However.1. then.2.1 ⎟ P 2 = ⎜ 2.
x 0 = (2.2) P' = (1. ) Note that all 3 of the axioms must be proved by use of the weak axiom of the revealed preferences.5 ⎟ 2 ⎠ ⎝ ⎝ 2 ⎠
Is this a consistent consumer? To solve the problem.
. On the contrary. 4 x2 > x0. when x1 is chosen
P1 x1 < P1 x 0 ==> 10 < 12 ==> x 0 > x1
P1 x1 = P1 x 2 = 10 x 1 is revealed preferred to x 2 then.1. derive all the expenditure levels P0x0=12 P0x1=12 P0x2=13 P1x0=12 P1x1=10 P1x2=10 P2x0=17 P2x1=17.
x1 is revealed preferred to x2). then can we derive the utility function behind these demand curves? Steps:
M 1.. xn )) where F is any monotonic transformation... P1 x1 ≥ P1 x 2 .... Such that there exists a class of utility functions F (U (x1 .xk-1>xk (x1>x2 notation meaning.. Negative Slutsky terms d.e. then if x1>x2. Application (Integrability) Recall in Chapter 10 page 269 Max x1x2 s.Pn . c. P1 x1 + P2 x 2 = M ==> x1M =
M M M .
a.. x2 ... then
P k x k < P k x1 .. P 2 x 2 ≥ P 2 x3 . x2>x3...
(Guarantees that I-curves do not intersect. Strong Axiom of the Revealed Preferences Let xi be purchased at Pi. x 2 are really the demand functions. are derivable from utility analysis. which. Check S11S22 – (S12)2 = 0 5
. M ) that are consistent with the strong axiom of revealed preferences. So we need another stronger axiom ⇒ The Strong Axiom of the Revealed Preferences. when maximized subject to the budget constraint ∑ Pi xi = M . Satisfy the budget constraint.. P k −1 x k −1 ≥ P k −1 x k . t. that is x k is not revealed preferred to x1 .. Check whether x1M .What is the reason for this? Actually the weak axiom does not imply that sij=sji (and the indifference curves are parallel).…. i..….) Now the main theorem: Theorem: Individual D – functions xi = xiM (Pi . Check S12 = S21 e. Homogenous of degree 0 b.. x2 = 2 P1 2 P2
Question: if these are the demand curves.. result in those particular demand functions.
tM M 0 M ⎪ = = x2 = t x2 ⎪ 2tP2 ⎭
M b) P1 x1M + P2 x 2 = P1
M M + P2 =M 2 P1 2 P2
c) S ij =
∂xiM ∂x M + xj i ∂Pj ∂M
∂x1M ∂x1M M ⎛ M ⎞⎛ 1 ⎞ M ⎟ ⎜ ⎟ S11 = + x1 = − 2 + ⎜ =− 2 <0 ⎜ ⎟ ⎜ ⎟ ∂P1 ∂M 2 P1 ⎝ 2 P1 ⎠⎝ 2 P1 ⎠ 4 P1
M M ∂x 2 M M ∂x 2 = + x2 =− 2 <0 ∂P2 ∂M 4 P2 M ∂x1M M ⎫ M ∂x1 + x2 = ⎪ ∂P2 ∂M 4 P1 P2 ⎪ ⎬S12 = S 21 M ⎪ = ...x1M =
tM ⎫ = x1M = t 0 x1M ⎪ 2tP1 ⎪ ⎬ both of them are homogenous to degree zero.......Then let us find the utility function behind Slope of Indifference Curve
dx2 MUx1 P = − MRS = − =− 1 dx1 MUx2 P2
dU = U 1 dx1 + U 2 dx 2 = 0 dx 2 U =− 1 dx1 U2
U1 P = 1 U 2 P2 ==> dx2 P =− 1 dx1 P2
..... = 4 P1 P2 ⎪ ⎭
d) S12 =
e) For negative semi definiteness
S11 S 22 − (S12 ) 2 = 0
M so x1M ........ x 2 exhibit all the properties......
But this might not be possible always with more than two variables.
nx 2 = −nx1 + nF (U ) nx 2 + nx1 = nF (U ) F (U ) = x1 x 2 ==> utility function
With this solution several things went right.
dx2 p h (x . P2 = 2 x1 2 x2
M P1 2 x1 x1 = = P2 M x2 2 x2
First order equation that have to be solved. x2 = 2 x1 2 x2
M M . M )
M p1 x1M + p2 x2 =M
M 2 M 2
Then if there exists a utility function behind. p2 . x2 )dx1 + h2 ( x1 . then s12 = s21 must be.
dx 2 x dx dx = − 2 ==> 2 = − 1 dx1 x1 x2 x1
Integrate both sides.x1M = P1 =
M M M . x2 )dx2 = 0 is a function not easy to solve always. But luckily for the two
variable case a solution always exists (by integrating factor method). p2 . M ) x = x ( p1 . x ) = − 1 = − 1 1 2 and dx1 p2 h2 ( x1 . Let’s check!!
∂x1M ∂x M ∂x M p1 + 1 p2 + 1 M = 0 ∂p1 ∂p2 ∂M
. In genereal. Now take
x1M = x1M ( p1 . For example slope of the indifference curve could be expressed in terms of x1 and x2. x2 )
h1 ( x1 .
then this utility function will have the usual convex I-curve of consumer theory.(3) ⇒ p1 (
∂x1M ∂x M ∂x M ∂x M + x1 1 ) + p2 ( 1 + x2 1 ) = 0 ∂p1 ∂M ∂p2 ∂M
⇒ p1s11 + p2 s12 = 0
Similarly we get
⇒ p1s21 + p2 s22 = 0
Now differentiating (3) w. M ) .r.
∂x1 ∂x + p2 2 = − x1 ∂p1 ∂p1
∂x1 ∂x + p2 2 = 1 ∂M ∂M
Then multiply (5) by − x1 to equate it to (4). it is always possible to find a utility function U which 2 generates D-curves xiM = xiM ( p1 . p2 . p1 and then w. s11s22 − s12 = 0. Remark: So.t.r. M we get the following two functions. ∂x ∂x ∂x ∂x − x1 p1 1 − x1 p2 2 = p1 1 + p2 2 ∂M ∂M ∂p1 ∂p1
⇒ p1s11 + p2 s21 = 0
and similarly ⇒ p1s21 + p2 s22 = 0
Thus (A) and (C) gives s12 = s21 . s22 < 0.
.t. for the two variable case. If in addition s11 .