Contents

I Preferences and choice 1

1 2 3

1 Preference relations 1.1 Utility Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Choice Rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2 Consumer choice 4 2.1 The consumption set . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.2 Competitive Budgets . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.3 Demands Functions and Comparative Statistics . . . . . . . . . . 7 2.3.1 Wealth eect . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3.2 Price eects . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3.3 Implications of homogeneity and Walras law for price and wealth eects . . . . . . . . . . . . . . . . . . . . . . . . 10 2.4 The Weak Axiom of Revealed Preference and the Law of Demand 11 2.5 The Weak Axiom of Revealed reference and the Law of Demand 11

Part I

Preferences and choice

1 Preference relations
X :={set of mutually exclusive possible alternatives } Denition 1.1. Preference (technically binary) relations are dened over the set X and denoted by 1. Strict preference relation : x y () x y but y 6 x 2. Indierence relation :x y () x y and y x Denition 1.2. The preference relation is rational1 if it satises : 1. Completeness : 8x, y 2 X we have that x y or y x or both. 2. Transitivit y: 8x, y, z 2 X , if x y and y z =) xz. Completeness and transitivity are very strong assumptions, think how dicult it is to compare between options we do not know in real life. Proposition 1.1. if is rational, then:
1 Also

known as weak order

1.

is both irreexive ( x 6 y)

x) and transitive ( if x

y and y

z =)

2. is reexive and transitive 3. If x y z =) x y

1.1

Utility Functions

Denition 1.3. A function u : X ! R is a utility function representing (preference relation), if 8x, y 2 X , x y () u(x) u(y ) Notice that a utility function representing is not unique, in fact, any strictly increasing function f : R ! R, v (x) = f (u(x)) is a new utility function representing the same . Properties that are preserved under monotonic transformations are called ordinal. Thus associated with u(.) is an ordinal property. Cardinal properties are preserved under any transformation, e.g. dierences in utility meaure between alternatives are cardinal properties. Proposition 1.2. can be represented by a utility function only if it is rational Proof. The proof shows that if there is u(.) that represents =) is transitive and complete Rationality is necessary, but not suucient. In fact2 , for to be representable by a function u(.) it requires the following additional assumtion: Denition 1.4. Continuity of rational preferences3 (weak order). Let (X, d) be a metric space. We say is continuous, if 8x 2 X the sets B (x) = {y 2 X : y y } and W (x) = {y 2 X : y x} are closed in (X, d). An alternative denition, which is closer to continuty of a function on Rn is Denition 1.5. Continuity of preferences an % bn , an ! a and bn ! b, then a % b Proposition 1.3. Let X be a nite set, if % on X is rational, then it can be represented by a utility function. Proof. Dene the lower contour set of x 2 X as - (x) = {y 2 X : x % y }, x is weakly preferred to any element in this set. - (x) 6= ; since x % x =) x 2- (x). Dene u(x) = |- (x)| , a function that counts the number of elements in - (x). - (y ) - (x) =) |- (y )| |- (x)| () u(y ) u(x) , since we know we are not dealing with innity
2 Think 3 See

of lexicographic preferences and the jump experienced by a sequence of bundles De la Fuente page 329

1.2

Choice Rules

Denition 1.6. Choice behavior es represented by a choice structure (B, C ()) over the set of options X. 1. B is a collection of non-empy subsets of X . 2. C () is a choice rule (correspondence) that chooses a nonempty set of elements C (B ) B, 8B 2 B. Example: X = {x, y, z }, B{{x, y }, {x, y, z }} 1. c1 ({x, y }) = {x}, c1 ({x, y, z }) = {x}. 2. c2 ({x, y }) = {x}, c2 ({x, y, z }) = {x, y }. Note: Choice rules are selecting sets, so be careful with notation such as x 2 c(B 0 ), because this implies x is an element of the chosen set. But c(B 0 ) could contain other elements. Denition 1.7. Weak Axiom of Revealed Preferences (WARP) A choice structure (B, C ()) satises WARP : If for some B 2 B with x, y 2 B , we have x 2 C (B ), then for any B 0 2 B with x, y 2 B 0 and y 2 C (B 0 ), we must also have x 2 C (B 0 ). Note: According to this denition, under B , x has been revealed as weakly preferred to y , hence for any other B 0 , if x and y are feasible, it cannot be that y is chosen but not x. Denition 1.8. Given a choice structure (B, C ()) the revealed preference relation % is dened by 1. x % y () there is some B 2 B s.t x, y 2 B and x 2 C (B ) . 2.

: If 9B 2 B s.t x, y 2 B , x 2 C (B ) and y 62 C (B ).

Denition 1.9. Using the previous dention, it is possible to restate WARP as: If x is revealed at least as good as y , then y cannot be revealed preferred to x. Example: c2 ({x, y }) = {x}, c2 ({x, y, z }) = {x, y }. According to dention 1.8 c2 ({x, y }) = {x} =) x y , but then c2 ({x, y, z }) = {x, y } =) y % x!! Clearly, C2 () does not satises WARP. Fact 1.1. We may have a choice rule that satises WARP but that does satises completness in the denition of rational preferences4 . Denition 1.10. Consider a rational preference %, and any non-empty B X . A preference maximizing behavior has the following choice structure: C (B, % ) = {x 2 B : x % y, 8y 2 B }.
4 Consider

C2 in example 1.C.1 of MWG or come up with your own idea !

In words, it is a choice structure where the underlying rule is %. Proposition 1.4. If % is rational =) (B, C (, %) satises WARP. Proof. Suppose 9B 2 B s.t. x, y 2 B and x 2 C (B, %), by the denition of C (B, %) , it is implied that x % y 2 B . To verify WARP holds, let B 0 2 B , with x, y 2 B 0 where y 2 C (B 0 , %). Hence, y % z , 8z 2 B 0 . Thus, by transitivity x % y % z =) x % z , 8z 2 C (B 0 , %) =) x 2 C (B 0 , %). In words, the proof states that given that the set C (B 0 , %) contains all elements that are weakly preferred to 8z 2 B 0 and x % y , it must be the case that x 2 C (B 0 , %). Notice that x must also be feasible. Proposition 1.5. If (B, C ()) is a choice structure5 that 1. WARP is satised. 2. B includes all subsets of X of op to three elements. Then, 1. Theres is a rational preference relations % that rationalizes C () relative to B; that is, C (B ) = C (B, %)8B 2 B. 2. Furthermore, this % is the only preference raltation that does so.

Consumer choice
1. Market economy : it is a setting in which the goods and services that the consumer may acquire are available for purchase at known prices. 2. Consumer: the most elementary decision making unit to the study of consumer demand. 3. Commodities: The objetct of choice. 4. Consumption set: The set of consumer choice subject to physical constraints. 5. Walrasian Budget Set: The set of consumer subject to physical AND economic constrain. 6. Comparative Statics: The study of Walrasian demand response to changes of exogenous parameters/constrains such as prices and wealth.
5 This

Some basic ideas and implicit dentions used freely in this chapter are:

is proposition 1.D.2 of MWG, the proof is on page 13.

2.1

3 x1 6 . 7 L Denition 2.1. Commodity vector ( or commodity bundle) x = 4 . . 52R . xL

The consumption set6

For simplicity we assume that the number of ommodities is nite and equal to L (indexed by l = 1m..., L). Denition 2.2. Consumption set: RL X = {x 2 RL : xl 1, 2, ..., L} An importan feature of our consumption set is that it is convex. Denition 2.3. Convex set (strict convex) if 8x, x0 2 A, x = x+(1 )x0 , 2 [0, 1] =) x 2 A (x = x + (1 )x0 , 2 (0, 1) =) x 2 A. 0, 8 l =

2.2

There are a few implicit assumptions throughout this section and they are: completeness of markets, price-taking behavior. Denition 2.5. The Walrasian or competitive budget set Bp,w = {x 2 RL + : p x w} is the set of all feasible consumption bundles for the consumer who faces market prices p and has wealth w. In sum, the consumers problem consists in choosing a consumption bumdel x 2 Bp,w , given p and w. Denition 2.6. The set {x 2 RL : p x = w} is called the budget hyperplane. The budget hyperplane determines the upper boundary of the budget set.
6 Check

3 p1 6 . 7 L Denition 2.4. Price vector p = 4 . . 5 2 R and p >> 0. pL

Competitive Budgets

page 19 of MWG to see a mix of continuous and discrete boundles and combinations.

The following graph shows the geometric relationship between p and the budget hyperplane.

x2

p w/p2

x={x'-x} x'

Bw,p

w/p1
x1

The price vector p is orthogonal to the budget hyperplane, because for any x0 that lies on it, we have p x0 = p x = w. Hence p x = 0 for x = (x0 x). Lemma 2.1. The walrasian budget set Bp,w is a convex set7 . Note: The budget constrain needs not be a straight line. Think of the consequences of dierent tax rates for dierents wages and the trade of eect of consumption and lesiure8 .

2.3

Demands Functions and Comparative Statistics

Denition 2.7. The consumers Walrasian ( or market, or ordinary ) demand correspondence x(p, w) assigns a set of chosen consumption bundles for each pair (p, w). In principle it could be multivalued, when this is so, any x 2 x(p, w) might be chosen. When x(p, w)is single-valued, we refer to it as a demand function. Denition 2.8. The Walrasian demand correspondence x(p, w) is homogenous of degree zero, if x(p, w) = x(p, w) for any p, w and > 0. This says that if prices and wealth change by , then consumption should stay the same.
7 see 8 page

proof on page 22 MWG 22 MWG if you cannot come with your own example

Denition 2.9. The walrasian demand correspondence x(p, w) satises Walras law if 8p >> 0 and w > 0, we have p x = w, 8x 2 x(p, w). Walras law says that the consumer fully expends his wealth. This assumption does not preclude savings. Fact 2.1. Througout this section, it is assume that the correspondence x(w, p) is homegeneous of degree zero and that it satises Walras law. Note: A convinient implication of being homegeneous of degree zero is that, although x(p, w) formally has L + 1 arguments, we can normalize the leel of one of this L + 1 independen variables at an arbitrary level. This reduces the efective level or arguments in x(p, w) to L. A common normalization is to set come pl = 1 w = 1. 2 3 x 1 ( p1 , w ) 6 x2 (p, w) 7 6 7 x(p, w) = 6 7 . . 4 5 . xL (p, w) 2.3.1 Wealth eect Denition 2.10. For a xed price p , the function of wealth x( p, w) is called consumers Engel function. Its image in RL p, w) : w > 0}, is known = {x( + , Ep as the wealth (income) expansion path. Wealth eect is dened as: @ xl (p, w)/@ w 1. If @ xl (p, w)/@ w 0 we say commodity l is normal. If every commodity is normal, we say the demand is normal. 2. If @ xl (p, w)/@ w < 0 we say commodity l is inferior. 6 6 In matrix notation the weatlh eects are reprsented as: Dw x(p, w) = 6 6 4 R .
L

@ x1 (p,w) @w @ x2 (p,w) @w

. . .

@ xL (p,w) @w

7 7 72 7 5

2.3.2

Price eects

Denition 2.11. Price eect is dened as @ xl /@ pk .9 In matrix notation it is: 2 @ x (p,w) 3 1 1 (p,w ) @ x@ @ p1 pL 6 7 .. 6 7 . 4 5 @ xL (p,w) @ xL (p,w) @ p1 @ pL
9 Generally

it is that @ xl /@ pk < 0, but for Gien goods @ xl /@ pk > 0

x1

Bp,w**

w**>w*>w

x(p,w**)

Ep

x(p,w*)

Bp,w* x(p,w)

x2

Figure 2.1: Shows the income (wealth) expansion path for a given price vector p.

p2

x2(p1,.,w) x2(p*1,.,w) x2(p**1,.,w)

x2

Figure 2.2: Demand of x2 for xed p1 , p 1 , p1

x2 w/p"2 d w/p'2

x(p1,p2",w) w/p2

x(p1,p2',w)

x(p1,p2,w)

w/p1

x1

Figure 2.3: An oer curve as p2 changes, holding p1 constant. 2.3.3 Implications of homogeneity and Walras law for price and wealth eects

Proposition 2.1. If the Walrasian demand function x(p, w) is homegeneous of degree zero, then 8p and 8w:
L X @ xl (p, w)

k=1

@ pk

pk +

@ xl (p, w) w = 0, l = 1, ...L @w

In matrix notation Dp x(p, w)p + Dw x(p, w) = 0 Note: The result follows inmediately from homogeneity of degree zero, simply dierentiate x(p, w) x(p, w) = 0 and then, set = 1. The previous result can be restated by a simple manipulation yielding:
L X

"l,k (p, w) + "l,w (p, w) = 0, l = 1, ...L

k=1 @ xl (p,w) pk w l (p,w ) where "l,k (p, w) = @ x@ pk xl (p,w) ,and "l,w (p, w ) = @w xl (p,w) are the elasticities of good l with respect to price pk and elasticity of good l with respect to wealth, respectively.

10

Proposition 2.2. If the Walrasian demand function x(p, w) satises Walras L P l (p,w ) law for all p and w: pl @ x@ + xk (p, w) = 0, k = 1, ..., L or in matrix Pk notation p Dp x(p, w) + x(p, w)T = 0T . Similarly,
l=1

Proposition 2.3. If the Walrasian demand function x(p, w) satises Walras L P (p,w) law, then for all p and w: pl @ xl@ = 1 or in matrix notation p Dw x(p, w) = w 1.
k=1

Note 1: Proposition 2.2 and 2.3 are also known as Cournot and Engel aggregation, respectively. Note 2: Proposition 2.2 and 2.3 can be derived by dierentiating p x(p, w) = w ( the equality comes from Walras law), with respect to the vector p and scalar w respectively.

2.4

The Weak Axiom of Revealed Preference and the Law of Demand

In this section, it is implied that x(p, w) is single valued, homogeneous of degree zero and satises Walras; law. Denition 2.12. The Walrasian demand function( see denition 2.9 and fact 2.1 to understand this) x(p, w)10 satises the weak axiom of revealed preference (WA) if the following property holds for any two price-wealth situations (p, w) and (p0 , w0 ) : If p x(p0 , w0 ) w and x(p0 , w0 ) 6= x(p, w) =) p0 x(p, w) > w0 .
0

In words, if for given prices and wealth (p, w), x(p0 , w0 ) is aordable and x(p , w0 ) 6= x(p, w), then it must be that the bundle x(p, w) is unnafordable at prices p0 and wealth w0 . Another interpretation is that our consumer reveals that he prefers x(p, w) over x(p0 , w0 ). This must be the case, because he could aord x(p0 , w0 ) with (p, w) yet he decided to not to get this bundle.

2.5

The Weak Axiom of Revealed reference and the Law of Demand

Denition 2.13. Slutsky wealth compensation w = p x(p, w), where p = (p0 p). In words, if our consumer choses x(p, w) and he experiences a price change to p0 , then the Slutsky wealth compensation is the minimum ammount necessary such that the orginal bundle x(p, w) is aordable. Whenever price changes are accompanied by a Slutsky wealth compensation, we refer to them as compensated price changes.
10 It

is assumed that x(p, w) is unique througout this section

11

x2

Bp',w'

x(p',w')

Bp",w" Bp",w"

x(p",w") x(p',w')

Bp',w'

x(p",w")

Label

x1 x1

x1

Bp',w' Bp",w"

x(p',w')

Bp",w"

x(p',w')

x(p",w") Bp',w'

x(p",w")

x2

x2

x1

Bp",w"

x(p'.w')

Bp',w'

x(p",w") x2

Figure 2.4: On the left panel, when x(p0 , w0 ) is aordable, x(p , w ) is not. On the right panel, under the dierent Walrasian Budget sets, only one of the options is available at a time. On the lower left panel The denition is satised with equality, x(p, w) is the preferred. On the lower right panel, our consumer chooses x(p, w) under both circumstances. On the bottom panel, our consumer violates WA ( Wearl Axiom) because under Bp0 ,w0 , x(p, w) is available and he does not chose it, similarly, under Bp,w , x(p0 , w0 ) is available , yet he does not chose it.

12

x2

w/p2 Bp,w

w'=p'.w(p,w)

w'/p2 Bp',w' x(p,w) Label

Bp',w

w'/p'1 w/p1 x1

w/p'1

Figure 2.5: Under Bp,w , x(p, w) is chosen. After a price reduction from p to p0 , real wealth increases, hence, according to our denition, w < 0 . The latter, in a way such that x(p, w) is still aordable, as shown in the graph. 0 Notice that the initial displacement is to the point w/p1 , but after slutskys 0 0 weatlh compensation, the nal intersect is w /p1 . Moreover, the displacement is parallel from Bp0 ,w to Bp,w .

13

Proposition 2.4. Suppose that x(p, w) (Walrasian demand function) is h.d.z (homogeneous of degree zero) and satises Walras Law. Then x(p, w) satises the WA () the following property holds: For any compensated price change from an initial situation (p, w) to a new price-wealth pair (p0 , w0 ) = (p0 , p0 x(p, w)), we have (p0 p) [x(p0 , w0 ) x(p, w)] 0, with strict inequality whenever x(p, w) 6= x(p0 , w0 ). The following proof will us this fact: Fact 2.2. p x(p0 , w0 ) w and x(p0 , w0 ) 6= x(p, w) =) p0 x(p, w) > w0 holds () 8(p, w) and 8(p0 , w0 ), we have p0 x(p, w) > w0 whenever p x(p0 , w0 ) = w and x(p0 , w0 ) 6= x(p, w)11 . Proof. of Proposition2.13 ( =) ) For x(p, w) = x(p, w), clearly, (p0 p) [x(p0 , w0 ) x(p, w)]) = 0. So, suppose x(p, w) 6= x(p, w). Expanding our expression (p0 p) [x(p0 , w0 ) x(p, w)] = p0 [(x(p0 , w0 ) x(p, w)] p [(x(p0 , w0 ) x(p, w)]. Consider the rst term, p0 [(x(p0 , w0 ) x(p, w)], given that p0 is a compensated price change (see denition2.13), we know p0 x(p, w) = w0 12 and by Walras Law, p0 x(p0 , w0 ) = w0 . Hence our rst term p0 [(x(p0 , w0 ) x(p, w)] = 0. For our second term, -p [(x(p0 , w0 ) x(p, w)], the Weak Axiom implies that 0 if p x(p, w) w0 and x(p, w) 6= x(p0 , w0 ) =) px(p0 , w0 ) > w, moreover, Walras law dictates that p x(p, w) = w. Thus, -p [(x(p0 , w0 ) x(p, w)] = (p x(p0 , w0 ) w ) < 0. ((= ) This direction uses fact 2.2: that the WA holds () it holds for all compensated price changes13 . The argument is by contradiction. By contradiction, Assume we have p0 x(p00 , w00 ) < w0 and p00 x(p0 , w0 ) < w00 . That is, both bundles are aordable under either price-wealth scheme. Now, Bp0 ,w0 and Bp00 ,w00 are both convex (see lemma 2.1). Clearly, x(p00 , w00 ) and x(p0 , w0 ) sit on the intersection of these sets (theyre both aordable in either case), using the fact that the intersection of two convect sets is convex: Choose 2 (0, 1) for which [p0 + (1 )p00 ] x(p0 , w0 ) = [p0 + (1 )p00 ] 00 x(p , w00 ) and denote p = p0 + (1 )p00 and w = [p0 + (1 )p00 ] x(p0 , w0 ) = [p0 + (1 )p00 ] x(p00 , w00 ). Then we have w0 + (1 )w00 > p0 x(p0 , w0 ) + (1 )p00 x(p0 , w0 ) (we assumed this in the rst line of the proof) . Thus, w0 + (1 )w00 > p0 x(p0 , w0 ) + (1 )p00 x(p0 , w0 ) = w = p x(p, w) = )p00 (x(p, w)

[p0 + (1

(Notice that the new x(p, w) does not have to be aordable under both cases, see graph below)
11 Notice the notation has been slighted reverted to accomodate for the following proof. In addition, this fact is only a special case of the denition. 12 This makes the initial bundle x(p, w ) aordable with (p0 , w 0 ). 13 It is also obviating the trivial case when WA holds with equality

)p00 ] x(p, w) = p0 x(p, w) + (1

14

x2

Bp',w'

x(p,w)

Bp,w

x(p'',w'')

Bp'',w''

x(p',w') x1

Figure 2.6: The graph shows howx(p0 , w0 ) and x(p00 , w00 ) are aordable under either budget. Then how it is possible to build a new walrasian budget set Bp,w . Finally, the selected bundle x(p, w) does not have to be available under both budgets. Thats where the contradiction comes from, because this new setting obeys WA. Therefore, either p0 x(p, w) < w0 or p00 x(p, w) < w00 . Suppose p00 x(p, w) < w . Then we have x(p, w) 6= x(p00 , w00 ), p x(p00 , w00 ) = w (compensated price change) and p00 x(p, w) < w0 .
00 14

Remark 2.1. A short notation for (p0 p) [x(p0 , w0 ) x(p, w)] 0 is p x 0. Which can be intepreted as a form of law of demnds: Demand and price move in opposite directions. Denition 2.14. (p0 p) [x(p0 , w0 ) as Compensated Law of Demand15 . x(p, w)] 0 p x 0 is also known

For the case in which x(p, w) is dierentiable propostion 2.4 has the following implications: Let (p, w) be our starting point and let dp be a dierential change in prices. dw = x(p, w) dp (analogous to ( w = x(p, w) p) is a dierential compensated price change. The dierential change in demand induced by this compensated price change is dx = Dp x(p, w)dp + Dw x(p, w)dw16 =) by using dw = x(p, w) dp (2.1)
14 The

i.e y =

other case is identical and shown on page 32 of MWG. Compensated law of demand there is no gien goods 16 For example, if y = f (x , ..., x ) its total dierential is dy = 1 n) =
15 Under @ f ( ) dx1 @ x1

+ ... +

@ f () dxm .For @ xm

@y dx1 @ x1

+ ... +

@y dxn @ xn

y = f (k, l) =) dy = f kdk + f ldl

15

x1

x2

Bp,w

w'=p'.x(p,w)

Bp',w' x(p',w) Bp',w x(p,w)

Al lo wa b

le

lo c

at

io n sf or

x(

p',

w' )u nd e

Bp,w
rW A

x(p,w) x2 x1

Figure 2.7: The left panel shows that under compensated price change and WA, the compensated law of demand holds, because our consumer will either chose x(p, w) or a point that was not available before. The right panel shows that under WA without compensated price change, our consumer is free to chose any point on the line Bp0 ,w (for L = 2) and it might be possible that p1 # and x1 # without additional restrictrions. that dx = Dp x(p, w)dp + Dw x(p, w)[x(p, w) dp] [Dp x(p, w) + Dw x(p, w)x(p, w)T ]dp Using form the last expresion into dp dx 0 =) dp [Dp x(p, w) + Dw x(p, w)x(p, w)T ]dp 0 The expression inside brackets is equivalent to 2 3 S11 (p, w) S1L (p, w) 6 7 . . .. . . S (p, w) = 4 5 . . . SL1 (p, w) SLL )p, w) where the (l, k )th entry is @ xl (p, w) @ xl (p, w) + xk (p, w) @ pk @w (2.4)

(2.2) (2.3)

(2.5)

slk (p, w) =

(2.6)

Denition 2.15. Equation 2.5 is also known as Slutsky of substitution matrix, Equation 2.6 is also known as substitution eect.

16

Remark 2.2. slk (p, w) measures the dierential change in the consumption of commodity l (substitution to or from other commodities) due to a dierenital change in the price of commodity k when wealth is adjusted so that the consumer can still just aord his original consumption bundle. To see the latter, note that the change in demand for good l if wealth is left unchanged is (@ xl (p, w)/@ pl ) dpk . For the consumer to be able to just aord his original consumption bundle welath must vary bu the amount xk (p, w)dpk (dw due to change in pk ). The eect of this weatlh change on the demand for good lis then (@ xl (p, w)/@ w) [xk (p, w)dpk ]. The sum of this elements is therefore exactly slk (p, w)dpk . The result of the previous derivations is: Proposition 2.5. If a dierentiable Walrasian demand function x(p, w) satises Walras law, homegeneity of degree zero and weak axiom, then at any (p, w), the Slutsky matrix17 S (p, w) satises vS(p,w)0, 8v 2 RL . Remark 2.3. An implication of being negative semidenite implies sll (p, w) 0. The substitution with respect to its own price si nonnegative18 . Proposition 2.6. Suppose the Walrasian demand function x(p, w) is dierentiable, homogeneous of degree zero and satises Walras law. Then p S (p, w) and S (p, w)p = 0 for any (p, w). Thus according to the previous denition S (p, w) is always singular ( rank less than L).

17 This 18

does not be symmetric under our current assumptions, but for L = 2 it holds.

Remark 2.4. Thus a good can only be Gien a (p, w) only if it is inferior. In particular sll (p, w) = @ xl (p, w)/@ pl ) + [@ xl (p, w)/@ w] xl (p, w) 0, if @ xl (p, w)/@ pl ) > 0 =) @ xl (p, w)/@ w < 0

17