Microeconomics ’x %⇤ y , there is some B 2 B such that Price E�ects: Gi�en Goods Walras’ law and Wealth e�ects - Engel

oods Walras’ law and Wealth e�ects - Engel ag- ferential change in consumption of
x, y 2 B and x 2 C(B)’. The derivative @xl (p, w)/@pk is known as gregation commodity l due to the change in price
Preference and Choice Relationships the price e↵ect of pk for the l-th good. Alt- If the Walrasian demand function x(p, w) of commodity k when wealth is adjusted
1 Preference and Choice hough is natural to think that increase in satisfies Walras’ law, then for all p and (compensated) so that the consumer can
Define the set of most preferred alterna- still a↵ord the original bundle.
Preference Relation tives in every B 2 B as, price lowers the demand, a Gi↵en Good is w:
Preference relations are binary relation XL
@x (p, w) Proof: Write the Law of de-
C ⇤ (B, %) = {x 2 B|x % y for every y 2 B} such that, @xl (p, w)/@pl > 0. Price e↵ects
pl l =1 mand as dp · dx  0, substitute
on the set of alternatives x, y 2 X: Suppose that % is a rational preference in matrix form are, @w
2 3 dx = Dp x(p, w)dp + Dw x(p, w)dw
• Weak Preference %: It is read "x is at relation. Then the choice structure gene- 66 @x1 (p,w) . . . @x1 (p,w) 77 l=1
least as good as y". 66 @p1 77 Or, to obtain the substitution matrix:
rated by %, (B, C(·, %)) satisfies the weak Dp x(p, w) = 666 @x (p,w) @x@p(p,w) L 77
pDw x(p, w)p = 1
• Strict Preference : ’x is preferred to y’ axiom. 4 L
... L 75 dp · [Dp x(p, w) + Dw x(p, w)x(p, w)T ]dp
or x y , x % y but not y % x. Proof: Suppose that x, y 2 B and x 2
@p1 @pL that is, the total expenditure must change Definiteness of Slutsky Matrix
• Indi↵erent relation ⇠: ’x is indi↵erent to by an amount equal to any wealth
C ⇤ (B, %), that is x % y. Suppose for B0 2 B Homogeneity on Price and Wealth e�ects change. If a di↵erentiable Walrasian demand
y’ or x ⇠ y , x % y and y % x. and x, y 2 B0 we have y 2 C ⇤ (B0 , %). This function x(p, w) satisfies the Walras’ law,
If the Walrasian demand function x(p, w) Proof: Simply di↵erentiate p · x(p, w) = w
Rational Preference Relation implies y % z, and since x % y by transiti- is homogeneous of degree zero, then for with respect to w. homogeneity of degree zero, and the
The preference relation % is rational if it vity x % z hence x 2 C ⇤ (B0 , %). all p and w: weak axiom, then at any (p, w), the Sluts-
possesses two properties: WARP and Walrasian Demand ky matrix S(p, w) satisfies v · S(p, w) · v  0
Rationalized choice rules XL
• Completeness: 8x, y 2 X we have x % y @xl (p, w) @x (p, w) The Walrasian demand function x(p, w)
Given a choice structure (B, C(·)), we say pk + l w=0 satisfies the weak axiom of revealed pre- for any v 2 RL (Slutsky matrix is negative
or y % x (or both). @pk @w semidefinite).
that the rational preference relation % k=1 ferences if the following property holds
• Transitivity: 8x, y, z 2 X if x % y and for l = 1, . . . L. Or This implies that the substitution e↵ect
rationalizes C(·) relative to B if C(B) = for any two price-wealth situations (p, w)
y % z, then x % z Dp x(p, w)p + Dw x(p, w)w = 0 of good l with respect to its own price
C ⇤ (B, %) for all B 2 B. and (p 0 , w0 ):
this implies: (i) is transitive and irrefle- This implies that changing by the sa- sl,l (p, w)  0 is always non positive.
xive (x x never holds). (ii) ⇠ is reflexive Existence of Rationalized choice if p · x(p 0 , w0 )  w and x(p, w) , x(p 0 , w0 ),
If (B, C(·)) is a choice structure such that, me proportion prices and wealth doesn’t 3 Classical Demand Theory
(x ⇠ x for all x), transitive, and symmetric change the demand. then p 0 · x(p, w) > w0 Monotonicity Assumption
(x ⇠ y then y ⇠ x). • the weak axiom is satisfied; Proof: By homogeneity of degree zero That is, if p · x(p 0 , w0 )  w and x(p, w) ,
• B includes all subsets of X of up to The preference relation % on X is mono-
Utility Function three elements; x(↵p, ↵w) x(p, w) = 0, di↵erentiating x(p 0 , w0 ) then we know that when facing tone if x 2 X and y >> x implies y x. It
A function u : X ! R is a utility function then there is a rational preference relati- with respect to ↵ and evaluating at ↵ = 1 (p, w) the consumer chooses x(p, w) even is strongly monotone if y x and y , x
representing % if 8x, y 2 X: yields the proposition above. though x(p 0 , w0 ) was a↵ordable. Hence, imply y x.
on % that rationalizes C(·) relative to B,
x % y , u(x) u(y) that is, C(B) = C ⇤ (B, %) for all B 2 B. Price and Wealth Elasticity x(p, w) is revealed preferred to x(p 0 , w0 ). This is also called desiderability assump-
For any strictly increasing function (mo- 2 Consumer Choice The previous proposition can be restated If at prices (p 0 , w0 ) the consumer chooses tion or the more is better assumption. Mo-
notonic increasing) f : R ! R, v(x) = in terms of elasticities. x(p 0 , w0 ), then it means that x(p, w) was notone preference imply that more of all
Walrasian Budget Set goods is preferred. Strongly monotone
f (u(x)) is a new utility function repre- The elasticity with respect to price: not a↵ordable.
The Walrasian, or competitive budget set imply that more of some good is prefer-
senting the same preferences as u(x). A @x (p, w) pk Slutsky compensation
preference relation % can be represented Bp,w = {x 2 RL⇤ |p · x  w} is the set of all "l,k (p, w) = l Price e↵ects have 2 e↵ects: (i) Change red.
@pk xl (p, w) Local nonsatiation
by a utility function only if it is rational. feasible consumption bundels for the con- the relative price, (ii) change the real we-
The elasticity with respect to wealth: The preference relation % on X is locally
Choice Rules sumer facing prices p = [p1 . . . pL ]0 and @x (p, w) w alth. Define the Slutsky wealth compensa-
A choice structure (B, C(·)) consists of: has wealth w. The set of commodities is "l,w (p, w) = l tion for a change in price from p to p 0 as nonsatiated if for every x 2 X and every
• B a family of nonempty subsets B of X RL⇤ = {x 2 RL |xl 0 for l = 1 . . . L}. @w xl (p, w) ✏ > 0, there is y 2 X such that ky xk  ✏
which give the percentage change in de- w0 = p 0 · x(p, w). Thus the bundle x(p 0 , w0 )
(B ⇢ X). B 2 B is called ’budget set’. The consumer problem can thus be stated is the bundle that the consumer would and y x.
• C(·) is a choice rule that assigns a non- ’Choose a consumption bundle x from Bp,w ’. mand for good l per percentage change in choose if wealth would not change from This is a weaker desiderability assump-
price of good k or wealth. So the previous
empty set of chosen elements C(B) ⇢ B Walrasian Demand: homogeneity proposition becomes, the situation p to the situation p 0 , so we tion: for every bundle x 2 RL+ within a
8B 2 B. The Walrasian demand correspondence L have isolated the e↵ect due to the change certain distance from x, there is another
For example X = {x, y, z}, B = X in relative prices.
x(p, w) = [x1 (p, w) . . . xL (p, w)] is homoge- "l,k (p, w) + "l,w (p, w) = 0 bundle y 2 RL+ s.t. y x.
{{x, y}, {x, y, z}} then a possible choice neous of degree zero if x(↵p, ↵w) = x(p, w) Law of Demand Convexity
structure (B, C1 (·)) is C1 ({x, y}) = {x} and for any p, w and ↵ > 0. k=1 For any compensated price change from
for l = 1. Thus an equal percentage The preference relation % on X is convex
C1 ({x, y, z}) = {x, z}. Walrasian Demand: Walras’ law an initial situation (p, w) to a new situati- if for every x 2 X the upper contour set
change in all prices and wealth leads to
Weak Axiom of Revealed Preferences The Walrasian demand correspondence no change in demand. on (p 0 , w0 ) = (p 0 , p 0 · x(p, w)), we have, {y 2 X : y % x} is convex. That is, if y % x
The choice structure (B, C(·)) satisfies x(p, w) satisfies Walras’ law if for every (p 0 p) · [x(p 0 , w0 ) x(p, w)]  0 and z % x, then ↵y + (1 ↵)z % x for any
Walras’ law and Price e�ects - Cournot
WARP, p >> 0 and w > 0, we have p · x = w for all
aggregation with strict inequality if x(p, w) , x(p 0 , w0 ). ↵ 2 [0, 1]. We refer to as diminishing mar-
’If for some B 2 B with x, y 2 B we ha- x 2 x(p, w). This law of demand tells that Demand ginal rates of substitution.
ve x 2 C(B), then for any B0 2 B with If the Walrasian demand function x(p, w) and price move in opposite directions.
Wealth E�ects - Normal & Inferior Goods satisfies Walras’ law, then for all p and Continuity
x, y 2 B0 and y 2 C(B0 ), we must also have At any (p, w), the derivative @xl (p, w)/@w Slutsky Substitution Matrix The preference relation % on X is con-
w:
x 2 C(B0 ).’ is known as the wealth e↵ect for the l-th XL Whenever a walrasian demand satisfies tinuous if for any sequence of pairs
In words, the weak axiom says that if x @x (p, w) the law of demand we can derive the L⇥L
good. pl l + xk (p, w) = 0 {xn , yn }1
n=1 with xn % yn for all n andx =
is ever chosen when y is available, there • If @xl (p, w)/@w 0 the good is called @pk Slutsky substitution matrix:
can be no budget set containing both x l=1  limn!1 xn , y = limn!1 yn we have x % y.
normal. s (p, w) . . . s (p, w)
and y for which y is chosen and x is not. for k = 1, . . . , L. Or, S(p, w) = s1,1 (p, w) . . . s1,L (p, w) No jumps in preferences.
• If @xl (p, w)/@w < 0 the good is called pDp x(p, w)p + x(p, w)T = 0T L,1 L,L
Utility Properties
Also, if x is not chosen when available, inferior. where the (l, k)-th entry is,
then x is never chosen when available. that is, the total expenditure cannot @x (p, w) @xl (p, 2) Monotonicity, Convexity and Continui-
In matrix form,
Revealed Preference h i0 change in response to a change in prices. sl,k (p, w) = l + xk (p, w) ty implies that the utility function u(·)
Given a choice structure (B, C(·)) the re- Dw x(p, w) = @x1 (p,w) . . . @x1 (p,w) Proof: Simply di↵erentiate p · x(p, w) = w @pk @w exists and that is increasing, quasi concave
vealed preference relation %⇤ is defined by @w @w with respect to all pk . where each sl,k (p, w) measures the dif- and di↵erentiable.
 Microeconomics objective function and constraint. We ha- welfare, namely equivalent variation and • If Production Set
ve, compensating variation. @x(p, w) Y ⇢ RL set of production vectors that are
• If x⇤ is optimal in UMP for w > 0, then <0
Equivalent Variation @w feasible for the firm. If y 2 Y then y is
UMP x⇤ is optimal in EMP for utility u(x⇤ ). then EV>CV. feasible, not otherwise;
Also, the minimized expenditure level Dollar amount that the consumer makes
Given p >> 0 and w > 0 the UMP is: a consumer indi↵erent to a price change EV/CV Which one? Transformation Function
in EMP is w;
v(p, w) = max u(x) s.t p · x  w
• If x⇤ is optimal in EMP for u > u(0), IF it had happened. Let u 0 = v(p 0 , w) and • EV takes old prices p 0 as reference; The transformation function F(Y ) is defi-
x 0
and if p >> 0 and u(x) is continuous, then then x⇤ is optimal in UMP when we- u 1 = v(p 1 , w) and e(p 0 , u 0 ) = e(p 1 , u 1 ) = • CV takes future prices p 1 . ned by Y = {y 2 RL |F(y) < 0}, and F(y) = 0
it has a solution. alth is p · x⇤ . Also, the maximized level w, then if it is not clear which one to use, it is if y is on the boundary of Y .
Walrasian Demand of utility in UMP is u. EV (p 0 , p 1 , w) = e(p 0 , u 1 ) e(p 0 , u 0 ) better to go for EV since it is in curren- Transformation Frontier
The solution of the UMP is the Walrasi- Duality 1: Indirect Utility/Expenditure = e(p 0 , u 1 ) w cy at current prices, and also it allows to The set of boundary points of Y is {y 2
an demand x(p, w). If u(·) is continuous it Function in terms of the indirect utility function,
compare more than one policy change.
RL |F(y) = 0}.
possesses: Duality allows to make an important 5 Aggregate Demand
• Homogeneity of degree 0: x(↵p, ↵w) = connection between the indirect utility v(p 0 , w + EV ) = u 1 Production Function
and the expenditure function: Aggregate Demand
x(p, w); If we distinguish in notation the set of
e(p, v(p, w)) = w and v(p, e(p, u)) = u Compensating Variation Aggregate demand is the sum of indivi- outputs from the inputs, we can define
• Walras’ law: p · x = w; dual’s demand functions:
• Unique: If u(x) is strictly quasiconcave, Compensation in dollars that returns the XI an ouput vector q = (q1 , . . . , qM ) 0 and
Duality 2: Hicksian/Walrasian consumer to the original utility after an an input vector z = (zM+1 , . . . ZL ) 0 note
then x(p, w) is unique. x(p, w1 , w2 , . . . , wl ) = xi (p, wi )
Duality allows to make an important economic change has occurred Let u 0 = that now inputs are non negative. Then
Indirect Utility Function connection between hicksian demand i=1 we can define a production function as,
The optimized value of the UMP, the in- and the walrasian demand: v(p 0 , w) and u 1 = v(p 1 , w) and e(p 0 , u 0 ) = thus it depends not only on prices q = f (z)
direct utility function, is defined as, h(p, u) = x(p, e(p, u)) e(p 1 , u 1 ) = w, then but also specific wealth levels. It would which is the maximum amount q of out-
v(p, w) = u(x(p, w)) and, be more
P convenient to write it in form
x(p, w) = h(p, v(p, w)) CV (p 0 , p 1 , w) = e(p 1 , u 1 ) e(p 1 , u 0 ) x(p, i wi ). Aggregate wealth cannot be
put produced by the inputs. If the good
where x(p, w) solves the UMP. That is, L is produced with inputs via the pro-
the utility function evaluated at the opti- = w e(p 1 , u 0 ) always computed, we need the property duction function, the production set is
mum. If u(·) is continuous, it has proper- Shepard’s Lemma in terms of the indirect utility function, that wealth expansion paths are parallel described as,
ties: If u(·) is continuous, then, v(p 1 , w CV ) = u 0 straight
P lines
P for 0every agent, hence for Y = {( z1 , . . . , zL 1 , q)|q f (z)  0}
• Homoegeneity of degree 0; h(p, u) = rp e(p, u) i=1 wi = i=1 wi then: where we use the minus for inputs for
• Strictly increasing in w, non increasing that is, EV/CV and hicksian demand & goods D(p, w1 , . . . , wI ) = D(p, w10 , . . . , wI0 ) convention.
in pl for all l; @e(p, u) This is formalized below. Properties of Production Sets
• Quasiconvex; hl (p, u) = Suppose that only p1 changes, and p l
@pl for all l. Then, since by Roy’s identity Aggregate Wealth • Y is nonempty: The firm must do some-
• Continuous is p adn w. thing;
EMP for all l = 1, . . . , L h1 (p, u) = @e(p, u)/@p1 : A necessary and sufficient condition for • Y is closed: The set contains its bounda-
Given p >> 0 and u > u(0) the UMP is: Slutsky Equation EV (p 0 , p 1 , w) = e(p 0 , u 1 ) w aggregate demand to be a function of ries;
Suppose u(·) is continuous, then, prices and aggregate wealth is that pre- • No free lunch: If y 2 Y and y 0 so that
e(p, u) = min p · x s.t. u(x) u
@hl (p, u) @xl (p, w) @xl (p, w) = e(p 0 , u 1 ) e(p 1 , u 1 ) ferences of all individuals admit indi-
x 0
= + xk (p, w) Z p0 rect utility functions of the Gorman form it contains no inputs. Then y = 0 that
1 @e(p, u 1 ) is output cannot be produced. Geome-
Hicksian Demand @pk @pk @w = dp1 with the coefficients on wi the same for
The Hicksian Demand h(p, u) is the so- for all l, k. Or in matrix notation, p11 @p1 every consumer i, i.e. trically, Y \ RL+ ⇢ {0}.
lution to the EMP. If u(·) is locally non Dp h(p, u) = Dp x(p, w)+Dw x(p, w)x(p, w)T Z p0
1
vi (p, wi ) = ai (p) + b(p)wi • Possibility of Inaction: 0 2 Y , so the firm
satiated, then: notice that this property implies, = h1 (p, u 1 )dp1 n.b. sufficiency is proved via roy’s identi- can shut down. If it is not possible,
• Homoegeneity of degree 0 in p: p11 ty. then there are sunk costs;
Dp h(p, u) = S(p, w) • Free disposal: If y 2 Y and y 0  y (so that
h(↵p, u) = h(p, u); similarly, Wealth Distribution Rule
• No excess utility: For any x 2 h(p, u), Roy’s Identity: Indirect to Walrasian y 0 produce at most the same amount of
CV (p 0 , p 1 , w) = e(p 1 , u 1 ) e(p 1 , u 0 ) When individual wealths are generated
output using at least the same amount
u(x) = u; Suppose u(·) is continuous and monoto- by a wealth distribution rule, that is, a
= e(p 0 , u 0 ) e(p 1 , u 0 ) of inputs), then y 0 2 Y i.e extra amount
• Convexity/uniqueness: If u(·) is strictly nic. Then, Z p0 set of functions (w1 (p, w) . . . wl (p, w)) with
quasiconcave, then h(p, u) is unique. rp v(p, w) 1 @e(p, u 0 ) PI of inputs (or outputs) can be disposed
x(p, w) = = dp1 i=1 wi (p, w) = w for all p, w, we can al- or eliminated;
Expenditure function rw v(p, w) p11 @p1 ways write the aggregate demand functi- • Irreversibility: If y 2 Y and y , 0 then
The optimized value of the EMP, the ex- that is, for every l = 1 . . . L: Z p0 on: y < Y . It is impossible to recoup the
1
penditure function, is defined has @v(p, w)/@pl = h1 (p, u 0 )dp1 D(p, w1 (p, w) . . . wi (p, w)) = D(p, w) inputs used to produce the output.
e(p, u) = p · h(p, u) xl (p, w) =
@v(p, w)/@w p11 • Non increasing returns to scale: For any
where h(p, u) solves the EMP. If u(·) is Properties of Aggregate Demand y 2 Y and ↵ 2 [0, 1] we have ↵y 2 Y .
continuous, e(p, u) has properties: Homoethetic Preferences Goods and EV/CV Aggregate demand satisfies: Any production vector can be scaled
• Homoegenity of degree one in p: A utility function v(x) is homothetic if The hicksian demand is not directly ob- • Homogeneity of degree 0; down.
e(↵p, u) = ↵e(p, u); v(x) = g(u(x)) where g is strictly increa- servable, but the walrasian demand (area • Walras Law; •
• Strictly increasing in u, non decreasing sing and u(x) is homogeneous of degree below hicksian) is. From this we note: The Weak Axiom need not to hold, we • Non decreasing returns to scale: For any
in pl for any l; 1. If preferences are homothetic: • If require that each walrasian demand satis- y 2 Y and ↵ 1 we have ↵y 2 Y . Any
• Concave in p; • e(p, u) = e(p)u; @x(p, w) fies the uncompensated law of demand. production vector can be scaled up.
• Continuous in p and u. • v(p, w) = v(p)w; =0 • Constant returns to scale: For any y 2 Y
@w 6 Production
Duality UMP/EMP • xi (p, w) = xi (p)w. then EV=CV. and ↵ 0 we have ↵y 2 Y . This is the
Production Vector
While UMP computes the maximal utility 4 Welfare Evaluation • If conjunction of increasing and decrea-
given w, the EMP computes the minimal This branch evaluates the e↵ects of chan- @x(p, w) The Production vector y = (y1 , . . . , yL ) 2 sing returns, i.e. any feasible produc-
>0 RL . If yl < 0 is input, if yl > 0 is output.
wealth to obtain u. The EMP is ’dual’ of ges in consumer welfare between t = 0 @w tion plan can be scaled up and scaled
UMP: captures the same aim reversing and t = 1: It requires a money metric for then EV<CV. Also known as production plan. down.
 Microeconomics • Maximize profits using the minimized Since Dp y(p) is symmetric and positive = {p 2 RN
+ |p1 + · · · + pN = 1}, here for and
costs. semidefinite, the law of supply holds in N = 3: {L00 % ↵L + (1 ↵)L0 }
even if ay seem an unnecessary complica- aggregate: if prices increase, so does sup-
• Additivity/Free entry: If y 2 Y and y 0 2 tion, the CMP is very useful as sometimes ply. (0, 0, 1) • Independence axiom: For all L, L0 , L00 2 L
Y , then y+y 0 2 Y . That is, two firms can we cannot well define the profit functi- Representative Producer and ↵ 2 (0, 1) we have,
set up feasible production plan and the on and the supply correspondence (e.g. Given Y1 , . . . , YJ we can define the aggre- L % L0 i↵ ↵L+(1 ↵)L00 % ↵L0 +(1 ↵)L00
may be not di↵erentiable, linear, or pro- that is, if we mix each lottery with
resulting production is y + y 0 , for this, gate production set as,
fits are +1) in which case the solution (0, 1, 0) (1, 0, 0) a third one L00 , the ordering of pre-
free entry must be possible. of the CMP and the value function are Y = Y1 + · · · + YJ
0
• Convexity: If y, y 2 Y and ↵ 2 [0, 1], better behaved. Let ⇡⇤ (p) and y ⇤ (p) be profit function ference between L and L0 does not
then ↵y + (1 ↵)y 0 2 Y . and the supply correspondence of the re- Each vertex stands for a degenerate lot- change. E.g. for ↵ = 1/2 then 12 L +
CMP tery where only one outcome is certain. 1 L00 % 1 L0 + 1 L00 i↵ L % L0 .
PMP The cost minimization problem can be presentative producer. Then if p >> 0: 2 2 2
P Each other point in the simplex represent
Given p >> 0 (price taking behavior) and stated as follows, • ⇡⇤ (p) = j ⇡j (p); a lottery with coordinate the probabili- Utility over Lotteries
P c(w, q) = min w · z and f (z) q If % satisfies the assumptions above, then
y 2 RL for p · y = Ll=1 pl yl , the profit ma- z 0 P ties.
• y ⇤ (p) = j yj (p); 7.1 Set of Feasible Lotteries there exist a utility function U : L ! R
ximization problem is, Conditional Factor Demand s.t.:
max p · y s.t F(y)  0 The solution to the CMP is z(w, q), it has hence, the aggregate profit obtained by The set of feasible lotteries is the set of L % L0 $ U (L) U(L0 )
y each individual firm that maximizes pro- simple lotteries: this is called Bernoulli utility
if Y corresponds to a single-output tech- properties: fits separately is the same as the maxi- L = {L1 , . . . , LN }
nology with production function f (z), • z(·) is homogeneous of degree 0 in w; Expected Utility
mum profit that can be obtained if all
then the problem can be restated as, • If f (·) is homogeneous of degree one firms merged and produced as a single Example If % satisfy continuity and independence,
max p · q w · z s.t q  f (z) then z(·) is homogeneous of degree 1 in firm. Play roulette and place ”100” with possi- then % admits a utility function u : X !
z 0 q; bility to win 3500 R such that,
or, E�icient Production N N
• If the set {z 0|f (z) q} is strictly con- X = { 100, 0, 3500} X X
max p · f (z) w · z A production vector y 2 Y is efficient if
vex, then z(w, q) is single-valued. 36 1 L % L0 $ pn u(xn ) pn0 u(xn )
z 0 there is no y 0 2 Y such that y 0 y and L = {( , 0, ), (0, 1, 0)}
with w being input prices. Cost Function y 0 , y. That is, a production vector is effi- 37 37 n=1 n=1
Supply Correspondence The optimized value of the CMP, the cost so the agent can maximize an expected
cient if there is no other feasible produc- Compound & Reduced Lottery utility E[u(x)] and not the utility over lot-
The solution to the PMP is y(p) and is a function c(w, q), if z(p, w) is single valued tion vector that generates as much output k k teries:
vector/set containing the optimal input- we can write, as y using no additional inputs, and that Given K simple lotteries Lk = (p1 , . . . , pN ) X
output demands that maximize profits. c(w, q) = w · z⇤ (w, q) produces more of some output or uses for k P = 1, . . . , K and probabilities ↵k 0 U(L) = max pn un (x)
x2X n
It is important to notice that not always The cost function is a value function that less of some input. with k ↵k = 1, the compound lottery
such a plan exists: a price system may be tells us what the minimal cost is at which (L1 , . . . , Lk ; ↵1 , . . . , ↵K ) is the risky alterna- This is called Von Neumann Morgenstern
First fundamental theorem of Welfare
such that there is no bound on how high the firm can produce quantity q at input utility.
If y 2 Y is profit maximizing for some tive that yields the simple lottery Lk with
profits may be, i.e. ⇡(p) = +1. Moreover, prices w. It has properties: probability ↵k , for k = 1, . . . , K.
p >> 0, then y is efficient. So profit ma- For any compound Not uniqueness of VNM utility
it can also be that it generates negative • c(·) is homogeneous of degree one in w lottery we can compu- Suppose that u : X ! R is a VNM utility
profits ⇡(p) < 0 in which case inaction is and non decreasing in q; ximization is sufficient condition for effi- te a corresponding reduced lottery with
ciency. representing %. Then, v : X ! R is ano-
allowed so that ⇡(p) = 0. It has proper- • c(·) is a concave function of w; the same final probability distribution ther VNM utility representing % if and
• If f (·) is homogeneous of degree one Second fundamental theorem of Welfare over outcomes. The value pn of the redu- only if exist a > 0 and b 2 R such that,
ties:
• y(p) is homogeneous of degree 0; then c(·) is homogeneous of degree 1 in Suppose that Y is convex. Then every effi- ced lottery is v(x) = a · u(x) + b
q; cient production plan y 2 Y is a profit- p = ↵ pn1 + · · · + ↵K pnK
• If Y is strictly convex, then y(p) is sin- maximizing production plan for some for n = 1,n. . . , N1. Therefore,
gle valued; • If f (·) is concave, then c(·) is a convex the reduced Lotteries over money
function of q; price vector p 0. That is, every produc- lottery is obtained by vector addition,
Notice that it does not have to be a signle tion plan can be decentralized to the op- Suppose the continuous variable x repres-
vector, but can be a set. Shepard’s Lemma timum by choosing the correct prices. L = ↵1 L1 + · · · + ↵K LK 2 ents an amount of money. Then for any x
Profit Function If z(w, q) consists of a single point, then we can define a cumulative distribution
7 Choice Under Uncertainty F : R ! [0, 1] s.t. F(x) = P(X  x), or if it
The profit function ⇡(p) is defined by c(·) is di↵erentiable with respect to w and, Outcomes
rw c(w, q) = z(w, q) admits pdf:
⇡(p) = p · y(p) X is the set of possible outcomes, someti- Z x
in case it is not single valued, it is a set, mes denoted C. It could be a set of possi- F(x) = f (t)dt
⇡(p) = max{p · y|y 2 Y } PMP: two stages ble consumption bundles or it could sim- 1
it has properties, Using the cost function, we can restate ply be the set of monetary payo↵s (X = R)
• ⇡(p) is homogeneous of degree 1; the firm’s problem (PMP) as, that the decision maker may receive. The
• ⇡(p) is convex; max pq c(w, q) elements of X are mutually exclusive and
q 0 exhaustive, i.e., for any feasible decisi-
Hotelling’s Lemma on one and only one element of X will
If y(p) consists of a single point, then ⇡(·) Aggregate Supply Expected Utility over Lotteries
The aggregate supply correspondence is eventually materialize. For simplicity we
is di↵erentiable at p and Preferences over Lotteries The expected utility from a lottery F(x) is
simply the sum of the individual supply assume that X is finite. The elements of The agent has preferences % over lotteries given by,
r⇡(p) = y(p) X are indexed by n 2 {1, . . . , N }.
correspondences. Assuming that yj (p) is with the following properties: E[u(x)] = u(x)dF(x)
that is, from the profit function we can Simple Lottery
derive its supply function. single-valued for all firms j 2 1, . . . , J we
A simple lottery LPis a list L = (p1 , . . . , oN ) • Complete: for any L, L0 2 L, L % L0 or
can write, L 0 % L; Risk Behavior
Cost Minimization with pn 0 and n pn = 1 where pn is
The cost minimization is a necessary con-
XJ • Transitive: for any L, L0 , L00 2 L with A decision maker is risk averter is a
the probability of outcome n occurring. degenerate lottery (lotteryRwith certain
dition for profit maximization, there is y(p) = yj (p) A sure consequence denoted xm is a de- L % L0 and L0 % L00 then L % L00 ;
no duality property. Thus, the PMP can j=1 generate lottery, that is, the lottery with • Continuity: for any L, L0 , L00 2 L and amount) that yields E[x] = xdF(x) with
also be solved in two necessary steps: the supply is not subject to wealth e↵ects, pm = 1 and pn = 0 for n , m. A sim- ↵ 2 [0, 1], the sets, certainty, is at least as good as the lottery
• Minimize cost; this simplifyies greatly the aggregation. ple lottery be represented in the simplex {↵L + (1 ↵)L0 % L00 } F(·) itself. If it is risk neutral, the agent is
 Microeconomics the foc for an optimum, agent j then riA > rjA . First order stochastic Dominance
q(1 ⇡)u 0 (w ↵q)+ Risk aversion and utility
+ ⇡(1 q)u 0 (w D + ↵(1 q))  0 The function r A (x) contains all relevant The distribution F(·) first order stocha-
indi↵erent between the two. If he is risk with equality if ↵ > 0. If the insurance
lover if the lottery F(·) is preferred to the price is actuarially fair then q = ⇡, thus information about u(x). Recall that, stically dominates G(·) if, for every non
degenerate lottery. d decreasing
Z function uZ: R ! R, we have,
the foc becomes, r A (x) = [ln u 0 (x)]
Jensen Inequality & Risk aversion u 0 (w ↵q) + u 0 (w D + ↵(1 q))  0 dx u(x)dF(x) u(x)dG(x)
The function u(x) is concave if and only if ↵ = 0 we would obtain u 0 (w D) < integrating,
Zx Z x
d first order stochastic dominance in terms
if, for all
Z F(·), Z ! u 0 (w) but since D > 0 it must be u 0 (w r A (t)dt = [ln u 0 (t)]dt of payo↵s, the distribution F(·) first order
D) > u 0 (w) hence the foc holds with equa- x x dt stochastically dominates G(·) if and only
u(x)dF(x)  u xdF(x) lity and ↵ > 0: = [ln u 0 (t)]xx if
if strictly concave the inequality holds u 0 (w ↵q) = u 0 (w D + ↵(1 q)) = ln u 0 (x) C F(x)  G(x)
strictly. Hence an agent is risk averse if by concavity, exponentiating both sides, for every x. If F(x) FOSD G(x)
u(x) is concave. w ↵q = w D + ↵(1 q) R x A then every expected utility maximi-
r (t)dt 0
Certainty Equivalent & risk aversion
hence, e x = e ln u (x) C = e C u 0 (x) zer whose utility function is strict-
↵⇤ = D integrating ly increasing prefers F(x) to G(x).
For a BNM u(·), the certainty equivalent thus with actuarially fair insurance the Z x R y Aup again, Z x
r (t)dt
of the lottery F(·) denoted c(F) is the agent insures completely. e x dy = e C u 0 (y)dy
amount of money for which the individu- Demand for Risky Asset x x
al is indi↵erent between the lottery F(·) Suppose a risk averse agent with wealth = e C [u(y)]xx
and the certain amount Z c(F, u), i.e. w can buy either of 2 assets:
= e C u(x) K
u(c(F, u)) = u(x)dF(x) • Safe asset with gross return (1 + r) = 1;
= au(x) + b
• Risky asset with gross return (1 + r) = z.
a decision maker is risk averter if for any the random return exceeds in mean the Arrow Pratt
F(·) the certainty equivalent is smaller safe asset, Z For two vN-M utility functions u1 (x) and
than the expected value Z of the lottery: E[z] = zdF(z) > 1 u2 (x) that are twice continuously di↵e-
c(F, u) < xdF(x) rentiable, strictly concave, and increa-
! denote ↵ and the amounts of we- sing, the following conditions are equiva-
Z
alth invested in the risky and safe assets lent:
u(c(F, u)) < u xdF(x) w = ↵ + . Thus, the portfolio or the new • r1A (x) r2A (x) an
Z Z ! wealth is, important implication of FOSD, is that
u(x)dF(x) < u xdF(x) w0 = ↵z + • u1 (x) is more concave than u2 (x), i.e. if F(x) first order stochastically domina-
the agent
Z maximizes E[w0 ], exists a function g(·) g 0 > 0 and g 00  0 tes G(x), then
Z Z
where the last holds, by the definition of such that u1 (x) = g(u2 (x))
c(F, u). u(↵z + )dF(z)s.t. ↵ + = w Relative Risk Aversion xdF(x) > xdG(x)
Risk Premium or substituting The relative risk aversion coefficient is,
Z the constraint, 00 but if the mean under F(x) bigger than
For a given VNM u(·), the risk premium r R (x) = x u (x) the mean under G(x) does not imply
max u(w + ↵(z 1))dF(z) u 0 (x) FOSD.
associated with F(·) denoted ⇢(F) is defi- 0↵w
ned by, ⇤
Z ! Z an ↵ solution
Z of the problem must (be, Types of VNM utility
 0 • Non-increasing absolute risk aversion Second Order Stochastic Domi-
u xdF(x) ⇢(F) = u(x)dF(x) (↵ ⇤ ) = (z 1)u 0 (w+↵ ⇤ (z q))dF(z)
0 (NIARA): if r A (x) is non-increasing in nance/Mean preserving Spread
which is equivalent that is, (↵ ⇤R)  0 if ↵ < W or (↵ ⇤ ) 0 if x;
Z to, • Decreasing absolute risk aversion
⇢(F) = xdF(x) c(F) ↵ > 0. Since zdF(z) > 1 if ↵ = 0 we have For any two distributions F(x) and G(x)
(DARA): if r A (x) is decreasing in x; with the same mean, F(x) second order
that byZthe KKT:
a decision maker is a risk averter, if and • Constant absolute risk aversion (CA- stochastically dominates (or is less risky
only if the risk premium for any lottery (0) = (z 1)u 0 (w)dF(z) RA): if r A (x) is constant in x; than) G(·) if for every nondecreasing con-
is positive. Z Z • Increasing absolute risk aversion (IA- cave function u : R+ !
Z Z R, we have:
7.2 Buy Insurance = u 0 (w) zdF(z) u 0 (w) dF(z) > 0 RA) if r A (x) is increasing in x.
Consider a risk averse decision maker u(x)dF(x) u(x)dG(x)
• Decreasing relative risk aversion
with wealth w who runs the risk of lo- Thus,↵ ⇤ = 0 cannot be a solution as it
doesn’t respect the KKT. Thus ↵ ⇤ > 0, that (DRRA): if r R (x) is decreasing in x; that is, F(x) is less ”risky” than G(x). An
sing D with probability ⇡. He can buy
insurance at price q per unit. Thus, if he is if a lottery is actuarially favorable, then • Constant relative risk aversion (CRRA): alternative interpretation is to say that
a risk averter will always want to accept if r R (x) is constant in x; F(·) is less risky than G(·) if G(·) is genera-
buys ↵ units of insurance his wealth is,
( a positive amount of it, no matter how • Increasing relative risk aversion (IA- ted by F(·) plus noise. To be more preci-
w1 = w q↵ w.p. 1 ⇡ no loss risk averse he is. se, suppose that G(x) describes the final
w2 = w q↵ D + ↵ w.p. ⇡ loss occurs RA) if r R (x) is increasing in x. probability distribution of a compound
Absolute Risk Aversion
thus, given VNM utility E[u(x)] and ber- The absolute risk aversion is, 7.3 Stochastic Dominance lottery. In the first stage, x is drawn ac-
noulli utility u(x) the maximization pro- u 00 (x) d Attaches meaning to the expression ”The cording to F(x). In the second stage, we
blem, r A (x) = = [ln u 0 (x)] distribution F(·) yields unambiguous- add to x the realization of a random va-
max(1 ⇡)u(w q↵) + ⇡u(w q↵ D + ↵) u 0 (x) dx ly higher returns than the distribution riable e " with E(e " |x) = 0. Thus, G(x) is a
↵ 0 if an agent i is more risk averse than G(·)”. mean-preserving spread of F(x).