Вы находитесь на странице: 1из 149

04/10/2017

EC202 2017/8
MICHAELMAS TERM SLIDE PACK

• five sections: Firm / Consumer / General Equilibrium / Risk & Uncertainty / Welfare
• first slide of each lecture has green border
• full versions at http://darp.lse.ac.uk/ec202

• for more, contact f.cowell@lse.ac.uk

THE FIRM:
Lectures 1 - 4
 Quantities  Functions
zi amount of input i  production function
z = (z1, z2 , , zm ) input vector C cost function
Z input requirement set Hi conditional demand for input i
q amount of output (single firm) S supply function
Di ordinary demand for input i
qf output of firm f
 Other
 Prices and profits  Lagrange multiplier (min cost)
wi price of input i  elasticity of demand
w = (w1, w2 , , wm) input-price vector
p price of output
 profits

1
04/10/2017

1-output, 2-input production function


q
 (z1, z2)
q > (z)
 Interior: feasible but inefficient
 Boundary: feasible and efficient

q =  (z) q < (z)  “Outside”: infeasible

z2
0

 In general: q   (z1, z2, , zm)


 Equivalently: q   (z)
  gives maximum output that can
be produced from given inputs

The
z
input requirement set
2

 Feasible but inefficient


Z(q)  Feasible and technically efficient
 Infeasible points

q < (z)
q = (z)

q > (z)

z1
 Pick a particular output level q
 Find a feasible input vector z
 Repeat to find all such vectors for given q
 Get the input-requirement set: Z(q) := {z: (z)  q}

2
04/10/2017

Case 1: Z smooth, strictly convex


 Pick two boundary points
z2
 Draw the line between them
 Intermediate points lie in the
interior of Z
Z(q)

 z
q = (z') q<  (z) Important role of convexity
Combination of two techniques
 z may produce more output

q = (z") What if we changed some of


the assumptions?
z1

Case 2: Z Convex (but not strictly)


 Pick two boundary points
z2
 Draw the line between them
 Intermediate points lie in Z
(perhaps on the boundary)
Z(q)

 z

 z
 A combination of feasible
techniques is also feasible
z1

3
04/10/2017

Case 3: Z smooth but not convex


 Join two points across the “dent”
z2
 Take an intermediate point

 Highlight zone where this can occur


Z(q)

 in this region there is an


indivisibility

z1

Case 4: Z convex but not smooth


z2

q = (z)
  Slope of the boundary is
undefined at this point

z1

4
04/10/2017

Isoquants
 Pick a particular output level q  Think of the isoquant as an

integral part of the set Z(q)


 Find the input requirement set Z(q)
 The isoquant is the boundary of
Z:{ z :  (z) = q }

 If the function  is differentiable at z  Where appropriate, use


then the marginal rate of technical subscript to denote partial
j (z) derivatives. So
substitution is the slope at z: —— (z)
i (z) i(z) := ——
zi .
 Gives rate at which you trade off
one input against another along the
isoquant, maintaining constant q

Isoquant, input ratio, MRTS


 The set Z(q)
z2  A contour of the function 
 An efficient point
 The input ratio
 Marginal Rate of Technical
Substitution
 Increase the MRTS
z2 / z1= constant
MRTS21=1(z)/2(z)  The isoquant is the
boundary of Z
 z′
 Input ratio describes one
z2°  z° production technique
{z:  (z)=q}
MRTS21: implicit “price”
of input 1 in terms of 2

z1  Higher “price”: smaller


z1° relative use of input 1

5
04/10/2017

MRTS and substitution


 Responsiveness of input ratio to MRTS

z2 z2

low high

z1 z1

Homothetic contours
 The isoquants
z2  Draw any ray through the origin…
 Get same MRTS as it cuts each
isoquant

z1
O

6
04/10/2017

Contours of a homogeneous function


 The isoquants
z2  Coordinates of input z°
 Coordinates of “scaled up” input tz°

tz2°  tz°

z2°  z° (tz) = tr (z)


trq
q
z1
O z1° tz1°

Case 1: IRTS
 An increasing returns to scale function
q  Pick an arbitrary point on the surface
 The expansion path…

  t > 1 implies (tz) > t(z)

z2
0

Double inputs and


you more than
double output

7
04/10/2017

Case 2: DRTS
 A decreasing returns to scale function
q  Pick an arbitrary point on the surface
 The expansion path…

 t > 1 implies (tz) < t(z)

z2
0

Double inputs and


output increases by less
than double

Case 3: CRTS
 A constant returns to scale function
q  Pick a point on the surface
 The expansion path is a ray


 (tz) = t(z)

z2
0

Double inputs and


output exactly doubles

8
04/10/2017

Relationship to isoquants
 Take any one of the three
q cases (here it is CRTS)
 Take a horizontal “slice”
 Project down to get the isoquant
 Repeat to get isoquant map

z2
0

Isoquant map is projection


of the set of feasible points

Marginal products

 Pick a technically efficient • Any z such that q= (z)


input vector

 Keep all but one input constant

 Measure the marginal change in • The marginal product


output w.r.t. this input
(z)
MPi = i(z) = ——
zi .

9
04/10/2017

CRTS production function again


q  Now take a vertical “slice”

 The resulting path for z2 = constant

z2
0

Let’s look at
its shape

MP for the CRTS function


(z)  The feasible set
q  Technically efficient points
(z)  Slope of tangent is the
marginal product of input 1
 Increase z1…

 A section of the
production function

Input 1 is essential:
If z1= 0 then q = 0

z1 1(z) falls with z1


(or stays constant) if
 is concave

10
04/10/2017

Relationship between q and z1


q q

 We’ve just taken the


conventional case

 In general this curve


depends on shape of 
z1 z1

 Other possibilities for


q
q
the relation between
output and one input…

z1 z1

Optimisation: the standard approach


 Choose q and z to maximise
m
 := pq –  wizi
i=1

 ...subject to the production


constraint...
• Could also write this as zZ(q)
q  (z)

 ..and some obvious constraints:


•You can’t have negative
q 0 z0 output or negative inputs

11
04/10/2017

A standard optimisation method


 If  is differentiable…
 Set up a Lagrangian to take care
of the constraints L (... )
 Write down the First Order 
 L (... ) = 0
Conditions (FOC) z

 Check out second-order 2


2L (... )
conditions z
 Use FOC to characterise solution z* = …

Stage 1 optimisation
 Pick a target output level q

 Take as given the market prices of inputs w

 Maximise profits...
m
 ...by minimising costs  wi zi
i=1

12
04/10/2017

Iso-cost lines
 Draw set of points where
z2 cost of input is c, a constant
 Repeat for a higher value
of the constant
 Imposes direction on the
diagram...
w1z1 + w2z2 = c"

w1z1 + w2z2 = c'

w1z1 + w2z2 = c
Use this to
z1 derive
optimum

Cost-minimisation
 The firm minimises cost...
z2
q  Subject to output constraint
 Defines the stage 1 problem
 Solution to the problem

minimise
m
 wizi
i=1

 subject to (z)  q
z* But solution depends on shape
of the input-requirement set Z

z1
What would happen in
other cases?

13
04/10/2017

Convex, but not strictly convex Z


z2

Any z in this set is


cost-minimising

z1  An interval of solutions

Convex Z, touching axis


z2

 Here MRTS21 > w1 / w2 at the


solution.
 z1  Input 2 is “too expensive”
z*
and so isn’t used: z2* = 0

14
04/10/2017

Non-convex Z
z2


z*
There could be multiple solutions.

But note that there’s no solution


point between z* and z**
z** 

z1

Non-smooth Z
z2

  z* is unique cost-
z* minimising point for q

True for all positive finite


z1 values of w1, w2

15
04/10/2017

Cost-minimisation: strictly convex Z


 Use the objective function
 Minimise  ...and output constraint
m
 wi zi
 ...to build the Lagrangian
q – (z)]
+ [q (z)  Differentiate w.r.t. z1, ..., zm ; set equal to 0
i=1  ... and w.r.t 

 Because of strict convexity we have  Denote cost minimising values by *

an interior solution
 A set of m+1 First-Order Conditions
* 1 (z* ) = w1
*2 (z*) = w2 
… … … 
*m(z *) = wm 
q = (z*)

From the FOC


 If both inputs i and j are used
and MRTS is defined then...
i(z) wi
——— = —
j(z) wj
 MRTS = input price ratio  “implicit” price = market price

 If input i could be zero then...


i(z) wi
———  —
j(z) wj
 MRTSji  input price ratio  “implicit” price  market price

16
04/10/2017

The solution
 Solving the FOC, get a cost-minimising value for each input...
zi* = Hi(w, q)
 ...for the Lagrange multiplier
* = *(w, q)
 ...and for the minimised value of cost itself
 The cost function is defined as
C(w, q) := min  wi zi
{(z) q}

Interpreting the Lagrange multiplier


 The solution function: At the optimum, either the
constraint binds or the Lagrange
C(w, q) =  iwi zi
*
multiplier is zero
= *
i wi zi – * [(z*) – q]
 Differentiate with respect to q: Express demands in terms of (w,q)
Cq(w, q) = iwiHiq(w, q)
– * [i i(z*) Hiq(w, q) – 1]
 Rearrange: Lagrange multiplier in (original) stage-
Cq(w, q) = i [wi – *i(z*)] Hiq(w, q) + * 1 problem is just marginal cost

Cq (w, q) = *

 Result is just an applications of a general “envelope” theorem


 It holds for the untransformed, original version of the problem
 If we use a transformed version of the constraint (for example log q  log (z))
we get a different Lagrange multiplier)

17
04/10/2017

Properties of C
z1 *  Draw cost as function of w1
C C(w, q+q)  Cost is non-decreasing in input prices

C(w, q)  Increasing in output, if  continuous


 Concave in input prices
°
 Shephard’s Lemma

C(tw+[1–t]w,q)  tC(w,q) + [1–t]C(w,q)

C(w,q)
w1 ———— = zj*
wj

What happens to cost if w changes to tw


 Find cost-minimising inputs for w, given q
z2 q  Find cost-minimising inputs for tw, given q

So we have:

• z*
C(tw,q) = i t wizi* = t iwizi* = tC(w,q)

The cost function is homogeneous


of degree 1 in prices.

z1

18
04/10/2017

Average and marginal cost


p increasing decreasing  The average cost curve
returns returns  Slope of AC depends on RTS
to scale to scale  Marginal cost cuts AC at its minimum

Cq
C/q

q
q

Revenue and profits


 A given market price p
 Revenue if output is q
 Cost if output is q
 Profits if output is q
 Profits vary with q
 Maximum profits
Cq
C/q

p

price = marginal cost

q q q q q q
q*

19
04/10/2017

What happens if price is low...

Cq
C/q

p price < average cost

q* = 0 q

Profit maximisation
 Objective: choose q to max

pq – C (w, q) “Revenue minus minimised cost”

 From FOC if q* > 0:


p = Cq (w, q*) “Price equals marginal cost”

C(w, q*)
p ———— “Price covers average cost”
q*
 In general:
covers both the cases:
pq*  C(w, q*)
q* > 0 and q* = 0

20
04/10/2017

The first response function


 Review the cost-minimisation
problem and its solution

 Choose z to minimise The “stage 1” problem


m
 wi zi subject to q  (z), z ≥ 0
i=1
 The firm’s cost function: The solution function
C(w, q) := min  wizi
{(z) q}

 Cost-minimising value for each input:  Hi is the conditional input


demand function
zi* = Hi(w, q), i=1,2,…,m
 Demand for input i, cond-
itional on given output level q

Mapping into (z1,w1)-space


Conventional case of Z
Start with any value of w1 ( the
slope of the tangent to Z)
Repeat for a lower value of w1
…and again to get…
z2 w1
…the conditional demand curve

  Constraint set is
convex, with smooth
boundary

 Response function is
a continuous map:

H1(w,q)
z1 z1

21
04/10/2017

Another map into (z1,w1)-space


Now take case of nonconvex Z
Start with a high value of w1
Repeat for a very low value of w1
Points “nearby” work the same way
But what happens in between?
A demand correspondence
z2 w1

 Constraint set is nonconvex



 Response is discontinuous
map: jumps in z*
 


 Map is multivalued at the

discontinuity

z1 z1

Use the cost function


 Recall this relationship?
Ci(w, q) = zi* Shephard's lemma

 So we have: Link between conditional input


Ci(w, q) = Hi(w, q) demand and cost functions

 Differentiate this with respect to wj Slope of conditional input


Cij(w, q) = Hji(w, q) demand function derived from
second derivative of cost function

22
04/10/2017

Two simple results


 Use a standard property second derivatives “commute”
2() 2()
——— = ———
wi wj wj wi

 So Cij(w, q) = Cji(w,q) The order of differentiation is irrelevant


Effect of price of i on conditional
 Therefore Hji(w,q) = Hij(w,q)
demand for j equals effect of the
price of j on conditional demand for i

 We also have: Cii(w, q) = Hii(w, q) We've just put j = i

 We know Cii(w, q)  0 Because cost function is concave

 Therefore: Hii(w, q)  0 The relationship of conditional


demand for an input with its own
price cannot be positive

Conditional input demand curve


w1  Consider the demand for input 1

 Consequence of result 2?
H1(w,q)

 “Downward-sloping”
conditional demand

 In some cases it is also


possible that Hii = 0

 Corresponds to the
case where isoquant is
kinked: multiple w values
consistent with same z*
z1

23
04/10/2017

The second response function


 Review the profit-maximisation
problem and its solution

 Choose q to maximise: The “stage 2” problem


pq – C (w, q)
 From the FOC:
p = Cq (w, q*), if q* > 0 “Price equals marginal cost”
pq*  C(w, q*) “Price covers average cost”

 Profit-maximising value for output:


q* = S (w, p)  S is the supply function

(again it may actually be a


correspondence)

Supply of output and output price


 Use the FOC: “marginal cost equals price”
Cq (w, q) = p

 Use the supply function for q:


Cq (w, S(w, p) ) = p Gives an equation in w and p

 Differentiate with respect to p Use the “function of a function” rule


Cqq (w, S(w, p) ) Sp (w, p) = 1

 Rearrange:
1 .
Sp (w, p) = ———— Gives slope of supply function
Cqq (w, q)

24
04/10/2017

The firm’s supply curve


p  The firm’s AC and MC curves
 For given p read off optimal q*
 Continue down to p
 What happens below p

Cq
Supply response is
C/q given by q = S(w,p)
 Case illustrated is
for  with first IRTS,
then DRTS
Response is a
_p – discontinuous map:
jumps in q*
| q Map is multivalued
_q at the discontinuity

The third response function


 Recall the first two response
functions:
zi* = Hi(w,q) Demand for input i,
conditional on output q

q* = S (w, p) Supply of output


 Now substitute for q* :
zi* = Hi(w, S(w, p) ) Stages 1 & 2 combined…
 Use this to define a new function:

Di(w,p) := Hi(w, S(w, p) ) Demand for input i (unconditional )

Use this relationship to


analyse further the firm’s
response to price changes

25
04/10/2017

Demand for i and the price of output


 Take the relationship
Di(w, p) = Hi(w, S(w, p))

 Differentiate with respect to p: Di increases with p iff Hi


increases with q. Reason? Supply
Dpi(w, p) = Hqi(w, q*) Sp(w, p) increases with price ( Sp > 0)

 But we also have, for any q: Shephard’s Lemma again


Hi(w, q) = Ci(w, q)
Hqi (w, q) = Ciq(w, q)

 Substitute in the above: Demand for input i (Di) increases


with p iff marginal cost (Cq)
Dpi(w, p) = Cqi(w, q*)Sp(w, p) increases with wi

Demand for i and the price of j


 Again take the relationship Di(w, p) = Hi(w, S(w, p))

 Differentiate w.r.t. wj: Dji(w, p) = Hji(w, q*) + Hqi(w, q*)Sj(w, p)

 Use Shephard’s Lemma again: Hqi(w, q) = Ciq(w, q)

 Use this and the previous result on Sj(w, p) to give a


decomposition into a “substitution effect” and an “output effect”:

Cjq(w, q*) “output effect”


Dji(w, p) = Hji(w, q*)   Ciq(w, q*)
“substitution
Cqq(w, q*) .

effect”

 Substitution effect is just slope of conditional input demand curve


 Output effect is [effect of wj on q][effect of q on demand for i]

26
04/10/2017

Results from decomposition formula


 Take the general relationship:
The effect wi on demand for
Ciq(w, q*)Cjq(w, q*)
Dji(w, p) = Hji(w, q*)   input j equals the effect of wj
Cqq(w, q*) . on demand for input i

 Now take the special case where j = i:


Ciq(w, q*)2
Dii(w, p) = Hii(w, q*)   If wi increases, the demand for
Cqq(w, q*) .
input i cannot rise

Input-price fall: substitution effect


w1  The initial equilibrium
 price of input falls
conditional demand
curve  value to firm of price fall,
given a fixed output level

H1(w,q)

Change in cost
price
fall

z1*
z1

27
04/10/2017

Input-price fall: total effect


w1  The initial equilibrium
 Substitution effect of input-
price of fall
 Total effect of input-price fall
price
fall

*
z1 z**
1 z1

The short-run problem


 Choose q and z to maximise
m
 := pq – wizi
i=1
subject to q  (z), q  0, z  0, and zm =zm

 Letq be the q for which zm =zm would be


chosen in the unrestricted problem
~ _
 C(w, q, zm ) := min  wi zi The solution function with
{zm =zm } the side constraint
 Short-run demand for input i:
~ _ ~ _
Hi(w, q, zm) =Ci(w, q, zm ) From Shephard’s Lemma
~ _
 C(w, q)  C(w, q, zm ) By definition of cost function
 So, dividing~by q: _
C(w, q)
______ C(w, q, zm )
q  _________
q
Short-run AC ≥ long-run AC

28
04/10/2017

MC, AC and supply in short and long run


 AC if all inputs are variable
 MC if all inputs are variable
 Fix an output level
p  AC if input m is now kept fixed
~
Cq  MC if input m is now kept fixed
 Supply curve in long run
~
Cq
 Supply curve in short run
C/q
C/q
SRAC touches LRAC at
 the given output

SRMC cuts LRMC at


the given output
q The supply curve is
q steeper in the short run

Conditional input demand


w1  The original demand curve for input 1

 The demand curve from the problem


H1(w,q) with the side constraint

 “Downward-sloping” conditional
demand

 Conditional demand curve is


steeper in the short run

~ _
H1(w, q, zm)
z1

29
04/10/2017

A market with two firms


 Supply curve firm 1 (from MC)
 Supply curve firm 2
 Pick any price
 Sum of individual firms’ supply
 Repeat…
 The market supply curve

p p p


q1 q2 q1+q2

low-cost high-cost both firms


firm firm

Market supply curve (2)


 Below p' neither firm is in the
market
 Between p' and p'' only firm 1
is in the market
 Above p'' both firms are in the
market
p p p


p" p"  
p' p' 
q1 q2 q1+q2

low-cost high-cost both firms


firm firm

30
04/10/2017

Take two identical firms…

p p

p' p'
q1 q2
4 8 12 16 4 8 12 16

Sum to get aggregate supply

p' •

8 16 24 32
q1 + q2

31
04/10/2017

Numbers and average supply

p  Rescale to get average supply of the firms


 Compare with S for just one firm
 Repeat to get average S of 4 firms
 …average S of 8 firms
 … of 16 firms

p' •••••••••••••••
average(qf)

4 8 12 16

The limiting case


The limit: continuous “averaged” supply curve
A solution to the non-existence problem?
A well-defined equilibrium
p Firms’ outputs in equilibrium

average
demand average
supply

p'

average(qf)

4 8 12 16

32
04/10/2017

Industry supply: negative externality


 Each firm’s S-curve (MC) shifted by the other’s output
 The result of simple MC at each output level
 Industry supply allowing for interaction

p p p
S1 (q2=5) S2 (q1=5) S
MC1+MC2

S1 (q2=1) S2 (q1=1)
MC1+MC2
q1
q2 q1+ q2

firm 1 alone firm 2 alone both firms

Analysing firms’ equilibrium


 price = marginal cost  determines output of any one firm

 price  average cost  determines number of firms


 Entry mechanism:
• if p  C/q gap is large then another firm could enter
• applying this iteratively determines the size of the industry

(0) Assume that firm 1 makes a positive profit


(1) Is pq – C ≤ set-up costs of a new firm?
• …if YES then stop. We’ve got the eqm # of firms
• …otherwise continue:
(2) Number of firms goes up by 1
(3) Industry output goes up
(4) Price falls (D-curve) and firms adjust output (individual firm’s S-curve)
(5) Back to step 1

33
04/10/2017

Firm equilibrium with entry


 Draw AC and MC
 Get supply curve from MC
 Use price to find output
marginal average  Profits in temporary equilibrium
price

cost cost  Allow new firms to enter

p
p
p 1
p  In the limit entry
 Price-taking
ensures profits are
temporary
competed away
p equilibrium
 p = C/q
 nf = 1234
output of
 nf = N
firm
qN q4q3qq21

Monopoly – model structure


 We are given the inverse demand function
• p = p(q)
• Gives the price that rules if the monopolist delivers q to the market
• For obvious reasons, consider it as the average revenue curve (AR)
 Total revenue is:
• p(q)q
 Differentiate to get monopolist’s marginal revenue (MR):
• p(q) + pq(q)q
• pq(·) means dp(·)/dq

 Clearly, if pq(q) is negative (demand curve is downward


sloping), then MR < AR

34
04/10/2017

Average and marginal revenue


AR curve is just the market
p demand curve…

Total revenue: area in the


rectangle underneath
Differentiate total revenue to get
marginal revenue

AR

MR
q

Monopoly – optimisation problem


 Introduce the firm’s cost function C(q)
• Same basic properties as for the competitive firm
 From C we derive marginal and average cost:
• MC: Cq (q)
• AC: C(q) / q
 Given C(q) and total revenue p(q)q, profits are:
• (q) = p(q)q  C(q)
 The shape of  is important:
• We assume it to be differentiable
• Whether it is concave depends on both C() and p()
• Of course (0) = 0
 Firm maximises (q) subject to q ≥ 0

35
04/10/2017

Monopoly – solving the problem


 Problem is “max (q) s.t. q ≥ 0,” where:
• (q) = p(q)q  C(q)
 First- and second-order conditions for interior maximum:
• q(q) = 0
• qq(q) < 0
 Evaluating the FOC:
• p(q) + pq(q)q  Cq(q) = 0
• p(q) + pq(q)q = Cq(q)
• MR = MC
 This condition gives the solution
• from above get optimal output q*
• put q* in p(∙) to get monopolist’s price
• p* = p(q* )

 Check this diagrammatically

Monopolist’s optimum
AR and MR
p
Marginal and average cost
Optimum where MC=MR
Monopolist’s optimum price

Monopolist’s profit
MC

AC

p* AR

MR
q
q*

36
04/10/2017

Monopoly – pricing rule


 Introduce the elasticity of demand :
•  := d(log q) / d(log p) = p(q) / qpq(q)
• <0

 First-order condition for a maximum:


• p(q) + pq(q)q = Cq(q)

 …can be rewritten as:


• p(q) [1+1/] = Cq(q)

 This gives the monopolist’s pricing rule:


Cq(q)
• p(q) = ———
1 + 

Monopoly – analysing the optimum


 Take the basic pricing rule
Cq(q)
• p(q) = ———
1 + 1/

 Use the definition of demand elasticity


• p(q)  Cq(q)
• p(q) > Cq(q) if | | < ∞
• “price > marginal cost”

 Clearly as | | decreases:


• output decreases
• gap between price and marginal cost increases

 What happens if | | ≤ 1 ( -1)?

37
04/10/2017

Monopolistic competition: 1
Take linear demand curve (AR)
The derived MR curve
Marginal and average costs
Optimal output for single firm
Price and profits
MC AC


 outcome is effectively
AR the same as for
monopoly
MR
output
of firm
q1

Monopolistic competition: 2

output
of firm
q1

38
04/10/2017

THE CONSUMER:
Lectures 5 - 8
 Quantities  Functions
xi consumption of good i U utility function
x = (x1, x2 , , xn ) consumption vector C cost (expenditure) function
X consumption set Hi compensated demand for good i
Ri resource stock of good i Di ordinary demand for good i
R = (R1,R2 ,,Rn) resource endowment V indirect utility function

 Prices and income  Other


pi price of good i  Lagrange multiplier (min cost)
p = (p1, p2 , , pn) price vector  Lagrange multiplier (max utility)
y money income  utility level

The budget constraint


A typical budget constraint
x2
Slope determined by price ratio
“Distance out” of budget line
fixed by income or resources

Two important cases determined by


1. … amount of money income y
2. …vector of resources R
p
– __1
p2
x1

39
04/10/2017

Case 1: fixed nominal income

x2
 Budget constraint determined by
the two end-points
  Examine the effect of changing p1
by “swinging” the boundary thus…

 Budget constraint is
n
 pixi ≤ y
i=1
y
__
.

p1

 x1

Case 2: fixed resource endowment


 Budget constraint determined by
x2 location of “resources” endowment R
 Examine the effect of changing p1 by
“swinging” the boundary thus…

 Budget constraint is
n n
 pi xi ≤  pi Ri
i=1 i=1
R

x1

40
04/10/2017

“Revealed Preference”
x2  Let market prices determine a person's budget
constraint
Suppose the person chooses bundle x…

 Use this to introduce Revealed Preference

 x′
x

x1

Axiom of Rational Choice Essential if observations


are to have meaning
Consumer always makes a
choice and selects the most
preferred bundle that is available

Weak Axiom of Revealed Preference


Weak Axiom of Revealed If x was chosen when x' was
Preference (WARP) available then x' can never be
chosen whenever x is available
If x RP x' then x' not-RP x

 Suppose that x is chosen when


prices are p
 If x' is also affordable at p then:

 Now suppose x' is chosen at


prices p'
 This must mean that x is not
affordable at p':

41
04/10/2017

WARP in action
x2
 Take the original equilibrium
 Now let the prices change…
WARP rules out some points
as possible solutions

 x°

 x′ Clearly WARP induces


a kind of negative
substitution effect
 x
 But could we extend
x1 this idea…?

Trying to extend WARP


x2 Take basic idea of revealed preference
Invoke revealed preference again
Invoke revealed preference yet again
 Draw the “envelope”

 x''

Is this an “indifference curve”…?

 x' No. WARP does not rule out


cycles of preference
 You need an extra axiom to
 x progress further on this
x1

42
04/10/2017

The (weak) preference relation


 The basic weak-preference "Basket x is regarded as at least
relation: as good as basket x' …"

x ≽ x'
 From this we can derive the “ x ≽ x' ” and “ x' ≽ x ”
indifference relation

x ~ x'
 …and the strict preference “ x ≽ x' ” and not “ x' ≽ x ”
relation…
x ≻ x'

Fundamental preference axioms


For every x, x'  X either x ≽ x' is true, or
 Completeness
x' ≽ x is true, or both statements are true

 Transitivity For all x, x', x"  X if x≽x' and x'≽x" then x ≽ x"

 Continuity For all x'  X the not-better-than-x' set and


the not-worse-than-x' set are closed in X
 Greed

 (Strict) Quasi-concavity

 Smoothness

43
04/10/2017

Continuity: an example
Take consumption bundle x°
x2  Construct two other bundles, xL
with Less than x°, xM with More
 There is a set of points like xL and
a set like xM
 Draw a path joining xL , xM
 If there’s no “jump”…
 xM
 x° But what about the
boundary points between
the two?
 xL
Do we jump straight from a
point marked “better” to one
marked “worse"?
x1

Utility function
 Representation Theorem:
• given completeness, transitivity, continuity
• preference ordering ≽ can be represented by a continuous utility
function
 In other words there exists some function U such that
• x ≽ x‘ implies U(x)  U(x')
• and vice versa
 U is purely ordinal
• defined up to a monotonic transformation
 So we could, for example, replace U(•) by any of the following
• log( U(•) )
• ( U(•) )
• φ( U(•) ) where φ is increasing
 All these transformed functions have the same shaped contours

44
04/10/2017

A utility function
  Take a slice at given utility level
 Project down to get contours

x2
0

Another utility function


  By construction U* = φ(U)
 Again take a slice…
 Project down …

x2
0

45
04/10/2017

The greed axiom


Pick any consumption
x2 bundle in X
Greed implies that these
bundles are preferred to x'
Gives a clear “North-East”
direction of preference

 Bliss!
B What can happen if
consumers are not greedy

 Greed: utility function is


monotonic
x' 

x1

Conventionally shaped indifference curves


x2 Slope well-defined everywhere
Pick two points on the same
indifference curve
Draw the line joining them
A  Any interior point must line on a
higher indifference curve

 ICs are smooth…


 C  ICs strictly concaved-contoured
I.e. strictly quasiconcave
B

(-) Slope is the Marginal


Rate of Substitution
x1 U1(x)
—— ..
U2 (x) .

46
04/10/2017

Other types of IC: Kinks


Strictly quasiconcave
x2
But not everywhere smooth

A

 C

B

x1

Other types of IC: not strictly quasiconcave

Slope well-defined everywhere


x2
Not quasiconcave
Quasiconcave but not strictly
quasiconcave

A

 C
B Indifference curves with flat
sections make sense
But may be a little harder to
work with…

x1

47
04/10/2017

The problem
 Maximise consumer’s utility U assumed to satisfy the
U(x) standard “shape” axioms

 Subject to feasibility constraint Assume consumption set X is the


xX non-negative orthant

 and to the budget constraint The version with fixed money


n income
 pixi ≤ y
i=1

The primal problem


 The consumer aims to maximise utility…
x2
 Subject to budget constraint
 Defines the primal problem
 Solution to primal problem

max U(x) subject to


n
 x*  pixi y
i=1

But there's another way


of looking at this
x1

48
04/10/2017

The dual problem


 Alternatively the consumer
x2 could aim to minimise cost…

 Constraint
set
 Subject to utility constraint
 Defines the dual problem
 Solution to the problem
 Cost minimisation by the firm

minimise
n
  pixi
x* i=1

subject to U(x)  

But where have we seen


x1
the dual problem before?

A lesson from the firm


 Compare cost-minimisation for the firm…
…and for the consumer

z2 q x2   The difference is
only in notation

 So their solution
functions and
response functions
must be the same
 
z* x*

z1 x1

49
04/10/2017

Cost-minimisation: strictly quasiconcave U


 Use the objective function
 Minimise …and utility constraint
n
 pi xi + λ[ – U(x)]
…to build the Lagrangian
U(x)  Differentiate w.r.t. x1, …, xn and set
i=1 equal to 0
 … and w.r.t 
 Because of strict quasiconcavity we  Denote cost-min values with a *
have an interior solution
 A set of n + 1 First-Order Conditions
 
 U1 (x ) = p1
 
 U2 (x ) = p2 
… … … 
 
 Un (x ) = pn 

 = U(x )

From the FOC


 If both goods i and j are purchased
and MRS is defined then…
Ui(x) pi
——— = —
Uj(x) pj
 MRS = price ratio  “implicit” price = market price

 If good i could be zero then…


Ui(x) pi
———  —
Uj(x) pj
 MRSji  price ratio  “implicit” price  market price

50
04/10/2017

The solution…

 Solve FOC to get a cost-minimising value for each good…

xi* = Hi(p, )
 …for the Lagrange multiplier
* = *(p, )

 …and for the minimised value of cost itself

 The consumer’s cost function or expenditure function is defined as


C(p, ) := min  pi xi
{U(x) }

Main results are immediate


 The cost function has same Same problem as for firm; so
properties as for the firm results are the same

 Shephard's Lemma gives


H is the “compensated” or
demand as a function of conditional demand function
prices and utility
H (p, ) = Ci(p, )
i

 Properties of the solution Downward-sloping with respect


function determine behaviour to its own price, etc…
of response functions

 “Short-run” results can be For example rationing


used to model side constraints

51
04/10/2017

Comparing firm and consumer


 Cost-minimisation by the firm…
 …and expenditure-minimisation by the consumer
 …are effectively identical problems
 So the solution and response functions are the same:

Firm Consumer
m n
 Problem: min wizi + [q – (z)] min pixi + [ – U(x)]
z i=1 x i=1

 Solution: C(w, q) C(p, )

 Response: zi* = Hi(w, q) xi* = Hi(p, )

The Primal and the Dual…

 There’s an attractive symmetry


about the two approaches to the n
problem  pixi+ [ – U(x)]
i=1

 In both cases the ps are given


n
and you choose the xs. But…
[
U(x) +  y –  pi xi ]
i=1
 …constraint in the primal
becomes objective in the dual…

 …and vice versa

52
04/10/2017

A useful connection
 Compare the primal
problem of the consumer…
…with the dual problem

x2 x2   Two aspects of the


same problem

 So we can link up
their solution functions
and response functions

 x* 
x*

x1 x1

Utility maximisation
 Use the objective function
 Maximise …and budget constraint
n …to build the Lagrangean
[
– pi xi
U(x) + μ y  ]  Differentiate w.r.t. x1, …, xn and
set equal to 0
i=1
 … and w.r.t 
 If U is strictly quasiconcave we have  Denote utility maximising
an interior solution values with a *

 A set of n+1 First-Order Conditions


U1(x ) = p1
U2(x ) = p2 
… … … 
Un(x ) = pn
n

y =  pi xi
i=1

53
04/10/2017

From the FOC


 If both goods i and j are purchased
and MRS is defined then…
Ui(x) pi (same as before)
——— = —
Uj(x) pj
 MRS = price ratio  “implicit” price = market price

 If good i could be zero then…


Ui(x) pi
———  —
Uj(x) pj
 “implicit” price  market price
 MRSji  price ratio

The solution
 Get U-maximising value for each good and Lagrange multiplier
• xi* = Di(p, y), i = 1,…,n
• * = *(p, y)
 Also for the maximised value of utility itself
 The indirect utility function is defined as
• V(p, y) := max U(x)
  pixi y}

• V is non-increasing in every price, decreasing in at least one price


• … is increasing in income y
• … is quasi-convex in prices p
• …is homogeneous of degree zero in (p, y)
• …satisfies “Roy's Identity”

54
04/10/2017

Another useful connection


 Indirect utility function maps The indirect utility function works
prices and budget into maximal like an "inverse" to the cost
utility: = V(p, y) function

 Cost function maps prices The two solution functions have


and utility into minimal to be consistent with each other.
budget: y = C(p, )

 Therefore we have: Odd-looking identities can be useful


= V(p, C(p, ))
0 = Vi(p,C(p,))+Vy(p,C(p,)) Ci(p,) Function-of-a-function rule
0 = Vi(p, y) + Vy(p, y) xi* Shephard’s Lemma

 Rearrange to get Roy’s identity The right-hand side is just Di(p, y)


xi* = – Vi(p, y)/Vy(p, y)

Utility and expenditure


 Utility maximisation
 …and expenditure-minimisation by the consumer
 …are effectively two aspects of the same problem
 So their solution and response functions are closely
connected:
Primal Dual
n n
 Problem:
x
[
max U(x) + μ y – pi xi
i=1
] min  pixi + [ – U(x)]
x i=1

 Solution: V(p, y) C(p, )

 Response: xi* = Di(p, y) xi* = Hi(p, )

55
04/10/2017

Solving the max-utility problem


 The primal problem and its solution
n
Lagrangean for the max U problem
max U(x) + [ y –  pi xi ]
i=1

U1(x*) = p1
U2(x*) = p2   The n + 1 first-order conditions,
… … …  assuming all goods purchased
Un(x*) = pn 
n

 pixi* = y
i=1

 Solve this set of equations:


x1* = D1(p, y)
x2* = D2(p, y)  Gives a set of demand functions, one
for each good: functions of prices and
… … …  incomes
xn* = Dn(p, y)
n
 A restriction on the n equations. Follows
pi Di(p, y) = y from the budget constraint
i=1

The response function


 Primal problem response function is Should be treated as just one
demand for good i: of a set of n equations
xi* = Di(p,y)

 The system has an “adding-up” Follows from budget


property: constraint: LHS is total
expenditure
∑ ,

 Each equation is homogeneous of  Again follows from the


degree 0 in prices and income. For budget constraint
any t > 0:
xi* = Di(p, y )= Di(tp, ty)

56
04/10/2017

Effect of a change in income

 Take the basic equilibrium


x2
 Suppose income rises
The effect of the income increase

 Demand for each good does not


fall if it is “normal”

 x**
 But could the opposite happen?

x*

x1

An “inferior” good
 Take same original prices, but
x2 different preferences
 Again suppose income rises
The effect of the income increase

 Demand for good 1 rises,


but…
 Demand for “inferior” good 2
falls a little

 Can you think of any goods


like this?

x*  x**
How might it depend on the
categorisation of goods?

x1

57
04/10/2017

Effect of a change in price


 Again take the basic equilibrium
x2
 Allow price of good 1 to fall
The effect of the price fall
The “journey” from x* to x**
broken into two parts

° x**


x*

x1

A fundamental decomposition
 Take the two methods of writing xi*:  Remember: they are two ways of
Hi(p,) = Di(p,y) representing the same thing
 Use cost function to substitute for y: Gives us an implicit relation in
Hi(p,) = Di(p, C(p,)) prices and utility

 Differentiate with respect to pj :  Uses y = C(p,) and function-of-a-


Hji(p,) = Dji(p,y) + Dyi(p,y)Cj(p,) function rule again

 Simplify : Using cost function and Shephard’s


Hji(p,) = Dji(p,y) + Dyi(p,y) Hj(p,) Lemma

= Dji(p,y) + Dyi(p,y) xj*  From the comp. demand function

 And so we get:
 This is the Slutsky equation
Dji(p,y) = Hji(p,) – xj*Dyi(p,y)

58
04/10/2017

The Slutsky equation


Dji(p,y) = Hji(p,) – xj* Dyi(p,y)
 Gives fundamental breakdown
of effects of a price change

 Income effect: “I'm better off if


the price of jelly falls; I’m worse
off if the price of jelly rises. The
size of the effect depends on
how much jelly I am buying…
 x**  …if the price change makes me
better off then I buy more normal
 x* goods, such as icecream”

 Substitution effect: “When the


price of jelly falls and I’m kept on
the same utility level, I prefer to
switch from icecream for dessert”

The Slutsky equation: own-price


 Set j = i to get the effect of the price of  Important special case
ice-cream on the demand for ice-cream

Dii(p,y) = Hii(p,) – xi* Dyi(p,y)


 Own-price substitution effect must be  Follows from the results on
negative the firm

 This is non-negative for normal goods  Price increase means less


disposable income
 So the income effect of a price rise must
be non-positive for normal goods

 Theorem: if the demand for i does not


decrease when y rises, then it must
decrease when pi rises

59
04/10/2017

Price fall: normal good


p1
 The initial equilibrium
ordinary
 price fall: substitution effect
demand curve
compensated  total effect: normal good

D1(p,y) (Hicksian)  income effect: normal good


demand curve
H1(p,)

For normal good income effect


must be positive or zero
price
fall

x*1 x**
1
x1

Consumer equilibrium: another view


x2
Type 2 budget constraint:
fixed resource endowment
Budget constraint with
endogenous income
 Consumer's equilibrium
Its interpretation

 Equilibrium is familiar:
same FOCs as before
 x*

 R
x1

60
04/10/2017

The offer curve


x2
 Take the consumer's equilibrium
 Let the price of good 1 rise
 Let the price of good 1 rise a bit more
 Draw the locus of points

 x***
 This path is the offer curve
 x**
 Amount of good 1 that household
supplies to the market
 x*

 R
x1

Household supply
 Flip horizontally , to make
supply clearer
 Rescale the vertical axis to
measure price of good 1
 Plot p1 against x1
x2 p1

 This path is the


 x***
household’s supply curve
 x** of good 1

 x*  The curve “bends back”


on itself
supply of
 R supply of good 1
Why?
good 1

61
04/10/2017

Decomposition – another look


 Take ordinary demand for good i: Function of prices and income
xi* = Di(p,y)
 Substitute in for y : Income itself now depends on
xi* = Di(p, j pjRj) prices

 Differentiate with respect to pj : The indirect effect uses


dx * dy function-of-a-function rule again
—i = Dji(p, y) + Dyi(p, y) —
dpj dpj
= Dji(p, y) + Dyi(p, y) Rj
 Now recall the Slutsky relation: Just the same as on earlier slide
Dji(p,y) = Hji(p,) – xj* Dyi(p,y)
 Use this to substitute for Dji: This is the modified Slutsky
dxi* equation
— = Hji(p,) + [Rj – xj*] Dyi(p,y)
dpj

The modified Slutsky equation:

dxi*
── = Hji(p, ) + [Rj – xj*] Dyi(p,y)
dpj

 Substitution effect has same interpretation as before


 Two terms to consider when interpreting the income effect
 This term is just the same as before
 This term makes all the difference:
• Negative if the person is a net demander
• Positive if the person is a net supplier

62
04/10/2017

The savings problem…


x2
 Resource endowment is non-
interest income profile
 Slope of budget constraint
increases with interest rate, r
 Consumer's equilibrium
 Its interpretation

 x1,x2 are consumption “today”


and “tomorrow”
 Determines time-profile of
consumption
 What happens to saving
 x* when the interest rate
changes…?

 R
x1

Labour supply…
x2
 Endowment: total time & non-labour
income
 Slope of budget constraint is wage rate
 Consumer's equilibrium

 x1,x2 are leisure and consumption

 Determines labour supply

 Will people work harder if their wage


rate goes up?
 x*
 R
x1

63
04/10/2017

The two aspects of the problem


 Primal: Max utility subject to
the budget constraint
 Dual: Min cost subject to a
utility constraint
 What effect on max-utility of
V(p, y) C(p,) an increase in budget?
 What effect on min-cost of
x2 x2 an increase in target utility?

C
V

 x* 
x*

x1 x1

Interpreting the Lagrange multiplier


 The solution function for the primal: At the optimum, either the
V(p, y) = U(x*)
= + [y –
U(x*) *  *
i pixi ]
constraint binds or Lagrange
multiplier is zero
 Differentiate with respect to y: Use the ordinary demand
Vy(p, y) = i Ui(x*)Diy(p, y) + * [1 – i piDiy(p, y)] functions
 Rearrange: Lagrange multiplier in the
Vy(p, y) = i[Ui(x*)–*pi]Diy(p,y)+* = * primal is MU of income

 The solution function for the dual: Same argument as above


C(p, ) = ipi xi* = ipi xi* – * [U(x*) – ]
 Differentiate with respect to  and rearrange: Lagrange multiplier in the
C(p, ) = i [pi–*Ui(x*)] Hi(p, )+* = * dual is MC of utility

 We can also show:. * = 1/ * A useful connection


between C and V

64
04/10/2017

The problem of valuing utility change

 Take the consumer's equilibrium


x2
'  and allow a price to fall...
  Obviously the person is better off.
...but how much better off?

x**


x*

x1

Story number 1 (CV)


 Price of good 1 changes
• p: original price vector
• p': vector after price change
 This causes utility to change
•  = V(p, y) original utility level at prices p
• ' = V(p', y) new utility level at prices p'

 Value this utility change in money terms:


• what change in income would bring a person
back to the starting point?
 Define the Compensating Variation:
original utility level restored
•  = V(p', y – CV)
at new prices p'
 Amount CV is just sufficient to undo
effect of going from p to p'

65
04/10/2017

The compensating variation

 A fall in price of good 1


x2
 Reference point is original utility level
  CV measured in terms of good 2

x**


x*

x1

Story number 2 (EV)


 Price of good 1 changes
• p: original price vector
• p': vector after price change
 This causes utility to change
•  = V(p, y) original utility level at prices p
• ' = V(p', y) new utility level at prices p'

 Value this utility change in money terms:


• what income change would have been needed
to bring the person to the new utility level?
 Define the Equivalent Variation:
new utility level reached at
• ' = V(p, y + EV)
original prices p
 Amount EV is just sufficient to mimic
effect of going from p to p'

66
04/10/2017

The equivalent variation


 Price fall as before
x2
'  Reference point is the new utility level
 EV measured in terms of good 2

x**


x*

x1

Welfare change as – (cost)


 Compensating Variation as –(cost): (–) change in cost of hitting utility
level . If positive we have a
CV(pp') = C(p, ) – C(p', ) welfare increase

 Equivalent Variation as –(cost): (–) change in cost of hitting utility


level '. If positive we have a
EV(pp') = C(p, ') – C(p', ') welfare increase

 Using these definitions we also have Looking at welfare change in the


reverse direction, starting at p'
CV(p'p) = C(p', ') – C(p, ') and moving to p

= – EV(pp')

67
04/10/2017

Cost-of-living indices
 An index based on CV: What's the change in cost of hitting the
C(p', ) base utility level ?
ICV = ————
C(p, )
 An approximation: What's the change in cost of buying the
i p'i xi base consumption bundle x?
IL = ———
i pi xi This is the Laspeyres index (the basis for
the Consumer Price Index)
ICV .
 An index based on EV:
What's the change in cost of hitting the
C(p', ') new utility level ' ?
IEV = ————
C(p, ')
 An approximation:
i p'i x'i
IP = ———
i pi x'i What's the change in cost of buying the
new consumption bundle x'?
IEV . This is the Paasche index .

Another (equivalent) form for CV


 Use the cost-difference definition:
CV(pp') = C(p, ) – C(p', ) (–) change in cost of hitting utility
level . If positive we have a
welfare increase
 Assume that the price of good 1
changes from p1 to p1' while other
prices remain unchanged. Then we
can rewrite the above as:
p1
CV(pp') =  C1(p, ) dp1 Using definition of a definite integral
p1'

 Further rewrite as:


p1
CV(pp') =  H1(p, ) dp1 Using Shephard’s lemma again
p1'

 CV is an area under the compensated


demand curve

68
04/10/2017

Compensated demand and the value of a price


fall (CV)
p1  The initial equilibrium
 price fall: (welfare increase)
compensated (Hicksian)
demand curve  value of price fall, relative to
original utility level

H1(p, )

The CV provides an exact


welfare measure
But it’s not the only approach
price

Compensating
fall

Variation

x*1
x1

Compensated demand and the value of a price


fall (EV)
p1 compensated  As before but use new utility
level as a reference point
(Hicksian)  price fall: (welfare increase)
demand curve
 value of price fall, relative
to new utility level

H1(p, )
The EV provides another exact
welfare measure
 But based on a different
reference point
Equivalent
price

Other possibilities…
fall

Variation

x**
1
x1

69
04/10/2017

Ordinary demand and the value of a price fall


 The initial equilibrium
p1 ordinary (Marshallian)  price fall: (welfare increase)
demand curve
 An alternative method of
valuing the price fall?

D1(p, y)
CS provides an approximate
welfare measure

Consumer's
price
fall

surplus

x*1 x**
1
x1

Three ways of measuring the benefits of a


price fall
p1

D1(p, y)

H1(p,) Summary of the three approaches.

H1(p, ) Illustrated for normal goods

For normal goods: CV  CS EV


price

For inferior goods: CV > CS > EV


fall

x1* x1**
x1

70
04/10/2017

GENERAL EQUILIBRIUM:
Lectures 9 - 12
 Quantities  Functions
qi aggregate net output of good i Uh utility function of h
xi aggregate consumption of good i f production function of firm f
Ri resource stock of good i Ei excess demand for good i
Rih resource holding by h of i
qi f net output by f of i  Other
x ih consumption by h of i N replication factor
1 2
[x ,x , …] allocation across households h reservation utiity for h
[q1,q2 , …] allocation across firms fh share of h in the profits of f
 Prices and incomes Q technology set
pi market price of good i A Attainable set
i shadow price of good i B “better than” set
f profits of firm f K Coalition
yh money income of h

Approaches to outputs and inputs


NET OUTPUT INPUTS
OUTPUTS A standard “accounting” approach
q1 z1 An approach using “net outputs”
How the two are related
q2 z2 A simple sign convention
... ...
qn-1 zm
qn q

q1 –z1 Outputs: +
net additions to the
stock of a good
q2 –z2 reductions in the
Inputs: 
... = ... stock of a good

qn-1 –zm Intermediate


0
your output and my
input cancel each
qn +q goods: other out

71
04/10/2017

The technology set Q


q1

Q
0

Tradeoff in inputs
q3

q4

(given q1 = 500)

(given q1 = 750)

72
04/10/2017

Tradeoff between outputs


q2 Again take slices through Q
 For low level of inputs
 For high level of inputs
 CRTS

high input

low input

q1

The technology set Q and the production function 

q2 A view of set Q: production possibilities of


two outputs.

(q) > 0 The frontier is smooth (many basic techniques)


Feasible but inefficient points in the interior
Feasible and efficient points on the boundary
Infeasible points outside the boundary
(q) = 0
 q  Q (q)  0

 (q1, q2,…,qn) nondecreasing in each qi


(q) < 0
Boundary is the transformation curve
Slope: marginal rate of transformation
MRTij := j (q) / i (q)
q1

73
04/10/2017

The Crusoe problem (1)


 max U(x) by choosing x and q  a joint consumption and
subject to... production decision
• xX
 logically feasible consumption

• (q)  0
 technical feasibility:
equivalent to “q Q ”
•xq+R
 materials balance:
can’t consume more of than is available
from net output + resources

Crusoe’s problem and solution


 Attainable set with R1= R2 = 0
x2  Positive stock of resource 1
 More of resources 3,…,n
 Crusoe’s preferences
 The optimum
 The FOC

 Attainable set derived from


technology and materials
• x* balance condition

 MRS = MRT:
U1(x) 1(q)
—— = ——
U2(x) 2(q)
x1
0

74
04/10/2017

Profits and income at shadow prices


 We know that there is no system of prices
 Invent some “shadow prices” for accounting purposes
 Use these to value national income

1q1 + 22q2 +...+


... nqn profits

1R1 + 2R2 +...+ nRn value of


resource stocks

1[q1+R1] +...+ n[qn+Rn] value of


national income

National income contours

q2+R2

1[q1+ R1] + 2[q2+ R2 ] = const

q1+R1

75
04/10/2017

“National income” of the Island

x2 Attainable set
 Iso-profit – income maximisaton
 The Island’s “budget set”
 Use this to maximise utility

U1(x) 
—— = —1
U2(x) 2  Using shadow prices 
 x* we’ve broken down the
Crusoe problem into a
1(x)  two-step process:
—— = —1
2(x) 2
1.Profit maximisation
2.Utility maximisation

x1
0

A separation result
 By using “shadow prices” … max U(x) subject to
 …a global maximisation problem xq+R
 …is separated into sub-problems: (q)  0

n
1. An income-maximisation problem max 
i=1
i [ qiRisubj. to
(q)  0

2. A utility maximisation problem max U(x) subject to


n
 i xi y
i=1

 Maximised income from 1 is used in problem 2

76
04/10/2017

Crusoe problem: another view


 The attainable set
 The “Better-than-x*” set
x2  The price line
 Decentralisation

 A = {x: x  q + R, (q)  0}

 B = {x: U(x)  U(x*)}


B
 x*
1  x* maximises income
 over A
2
 x* minimises expenditure
A over B
x1
0

Optimum cannot be decentralised


x2
 A nonconvex attainable set
 x°  The consumer optimum
 Implied prices: MRT=MRS
 Maximise profits at these prices

 x*
Production responses do not
A support the consumer optimum
 In this case the price system “fails”
x1

77
04/10/2017

Crusoe's island trades


x2  Equilibrium on the island
**
q2 **  q  The possibility of trade
 Max national income at world prices
 Trade enlarges the attainable set
 Equilibrium with trade

 x* is the Autarkic equilibrium:


 x* x1*= q1*; x2*=q2*
x2**
Domestic  x** World prices imply a revaluation
prices
World of national income
prices
In this equilibrium the gap
between x** and q** is bridged by
imports & exports

q1** x1** x1

The nonconvex case with world trade


x2
 Equilibrium on the island
 World prices
q2**  q**  Again maximise income at world prices
 The equilibrium with trade
 Attainable set before and after trade

x2**  x** Trade “convexifies” the


attainable set

 x*

A A′
x1
q1** x1**

78
04/10/2017

What is an economy?
 Resources R1 , R2 ,… n of these
(stocks)
 Households U1, U2 ,… nh of these
(preferences)

 Firms  ,… nf of these


(technologies)

An allocation
A competitive allocation consists of:
utility-maximising
 A collection of bundles (one for [x] := [x1, x2, x3,… ]
^
each of the nh households)
profit-maximising
 A collection of net-output vectors
^
(one for each of the nf firms) [q] := [q1, q2, q3,… ]

 A set of prices (used by


households and firms) p := (p1, p2, …, pn)

79
04/10/2017

How a competitive allocation works


 Implication of firm f’s profit
maximisation
 Firms' behavioural responses
p  {qf(p) , f=1,2,…,nf } map prices into net outputs
 Implication of household's
utility maximisation
 Households’ behavioural
responses map prices and
p, {yh } {xh(p) , h=1,2,…,nh } incomes into demands
 The competitive allocation

What does household h possess?


Rih  0,
i =1,…,n
 Resources R1 h , R2 h , …

0  fh  1,
f =1,…,nf

 Shares in firms’ 1h, 2h, …


profits

80
04/10/2017

Incomes
 Resources
 Shares in firms
Rents
 Net outputs
 Prices

Profits

The fundamental role of prices


 Net output of i by firm f depends  Supply of net outputs
on prices p:
qif = qif(p)

 Thus profits depend on prices:  Again writing profits as price-


n weighted sum of net outputs
f(p):=  pi qif(p)
i=1

 So incomes can be written as:


n nf
yh =  pi Rih +  fh  f(p)  Income = resource rents + profits
i=1 f=1

 Income depends on prices :  yh(•) depends on ownership


yh = yh(p) rights that h possesses

81
04/10/2017

Prices in a competitive allocation


 The allocation as a collection of
responses

p  {qf(p) , f=1,2,…,nf }  Put the price-income relation


into household responses
 Gives a simplified
p,p{yh } {xh(p) , h=1,2,…,nh } relationship for households

 Summarise the relationship

[q(p)]
p 
[x(p)]

The price mechanism


resource distribution  System takes as given the
share
a ownership
a property distribution
R1 , R2 , …  Property distribution consists
 1a,  2a, … of two collections
R1b, R2b, …  Prices then determine incomes

 1b,  2b, …  Prices and incomes determine


net outputs and consumptions

a
…  Brief summary…

d distribution
prices
[y]
allocation
[q(p)]
[x(p)]

82
04/10/2017

What is an equilibrium?
 What kind of allocation is an equilibrium?

 Again we can learn from previous presentations:


• must be utility-maximising (consumption)
• must be profit-maximising (production)
• must satisfy materials balance (the facts of life)

 We can do this for the many-person, many-firm case

Competitive equilibrium: basics


 For each h, maximise
n
Uh(xh), subject to  pi xih  yh
 Households maximise utility,
given prices and incomes
i=1  Firms maximise profits, given
prices

 For each f, maximise  For all goods the materials


balance must hold
n

pi qif, subject to f(qf )  0


i=1

 For each i:
xi  qi + Ri

83
04/10/2017

Consumption and net output


 “Obvious” way to aggregate  Appropriate if i is a rival good
consumption of good i?  Additional resources needed for each
nh additional person consuming a unit of i
xi = xih
h=1

 An alternative way to  Opposite case: a nonrival good


aggregate:  Examples: TV, national defence…

xi = max {xih }
h

 Aggregation of net output:  if all qf are feasible will q be feasible?


nf  Yes if there are no externalities
qi :=  qif  Counterexample: production with
f=1 congestion…

Competitive equilibrium: summary


 It must be a competitive  A set of prices p
allocation  Everyone maximises at those
prices p

 The materials balance  Demand cannot exceed supply:


condition must hold x≤q+R

84
04/10/2017

Alf’s optimisation problem


 Resource endowment
 Prices and budget constraint
x2a
 Preferences
 Equilibrium
R2a  Ra

 x*a  Budget constraint is


2 2
 pi xia ≤  pi Ria
i=1 i=1

 Alf sells some endowment of 2


for good 1 by trading with Bill
Oa R1a x1a

Bill’s optimisation problem


 Resource endowment
 Prices and budget constraint
x2 b
 Preferences
 Equilibrium

 x*b  Budget constraint is


2 2
 pi xib ≤  pi Rib
R2b  Rb i=1 i=1

 Bill, of course, sells


good 1 in exchange for 2
Ob R1b x1b

85
04/10/2017

Combine the two problems


x1b R1b Ob
 Bill’s problem (flipped)
 Superimpose Alf’s problem
x2a  Price-taking trade moves
agents from endowment point…
…to the competitive equilibrium
R2a  [R] R2b allocation
 The role of prices

 [x*]
 This is the Edgeworth box
 Width: R1a + R1b
 Height: R2a + R2b
x2b

Oa R1a x1a

Response to changes in prices


x2a  Alf’s endowment
 Alf’s reservation utility

Ra •  Alf’s preference map


 No trade if p1 is too high
R2a
 Trades offered as p1 falls
 Alf’s offer curve

• •

• •• • • •

R1a x1a
Oa

86
04/10/2017

Response to changes in prices (2)


ObBill’s situation…
x1 b x2bR1b …as an Australian

R2b  No trade if p1 is too low

Rb•
 Trades offered as p1 rises
 Bill’s offer curve







• •
Rb
R2b •
• R1xb 2b x1b
Ob •

Edgeworth Box and CE


x1b Ob (property
 Endowment point
distribution)
x2a R1b  The two offer curves
 Offers consistent at intersection
[R]•  By construction this is CE
R2a
R b U-maximising Alf
 Price-taking
2
 Price-taking U-maximising Bill
 Satisfies materials balance

• [x*]

x2b
R1a
Oa x1a

87
04/10/2017

Coalitions
 The population…
Viewed as nh separate individuals
 A coalition K…

 …is formed by any subgroup

K1

K2
K0

A formal approach

An allocation is blocked by a coalition if the


coalition members can do better for themselves

88
04/10/2017

Equilibrium concept
 Use the idea of blocking to introduce a solution concept
• if allocation is blocked a coalition could stop it happening
• such an allocation could not be a solution to the trading game
 So we use the following definition of a solution:
• the Core is the set of unblocked, feasible allocations

 Let’s apply it in the two-trader case


• In a 2-person world there are few coalitions:
{Alf }
{Bill}
{Alf & Bill}
• let’s see what allocations are blocked by them…
• …and what remains unblocked

The 2-person core


x1b b
x1 Ob
 Alf’s reservation utility
b
x
x2a  {Alf} blocks2these allocations
 Bill’s reservation utility
• [R]  {Bill} blocks these allocations
 Draw the contract curve
a  {Alf, Bill} blocks these allocations
• [xb]  The resulting core
 Bill gets all the advantage from
trade at this extreme point
a
x2 • [xa] b  Alf gets all the advantage from
trade at this extreme point

The contract curve


is the locus of
common tangencies

Oa a x2b
x1 x1a

89
04/10/2017

The core and CE


x1b
Ob
 The endowment point
x2a  The 2-person core again
 Competitive equilibrium again
• [R]

• [x*]  A competitive
equilibrium must
always be a core
allocation

x2b
Oa
x1a

The core and CE (2)


x1b
Ob curves that yield
 Indifference
multiple equilibria
x2a  Endowment point and
• [R]
reservation utility
 Equilibrium: low p1/p2
 Equilibrium: high p1/p2
a •  The core
[x*] b
• [x**]

x2b
Oa
x1a

90
04/10/2017

A simple result… and a question


 Every CE allocation must belong to the core Core
 It is possible that no CE exists
 What about core allocations which are not CE? CE
• Remember we are dealing with a 2-person model
• Will there always be non-CE points in the core?
 To find out, let's clone the economy
• economy replicated by a factor N, so there are 2N persons
• start with N = 2
• Alf and twin brother Arthur have same preferences and endowments
• likewise the twins Bill and Ben
 Now there are more possibilities of forming coalitions
• so more blocking!
{Alf} {Arthur} {Alf & Arthur} {Bill&Ben}
{Bill} {Ben} {Alf, Arthur &Bill} {Bill, Ben &Alf}
{Alf & Bill} {Arthur & Ben} {Alf, Arthur &Ben} {etc, etc}

Effect of cloning on the core

The core in the 2-person case


 [R] The extremes of the two-person core
{Alf,Arthur,Bill} can block [xa]…
 …leaving the Ben twin outside the
coalition

°
 [xb]
 Are the extremes still core
allocations in the 4-person
economy?
 [xa]
This new allocation is not a
solution…
But it shows that the core
must have become smaller

91
04/10/2017

How the blocking coalition works


Consumption in the coalition
Sum to get resource requirement

Consumption out of coalition

Alf xa = ½[xa+Ra]
 The consumption within
Arthur xa = ½[xa+Ra] the coalition equals the
coalition’s resources
Bill [2Ra +Rb – 2xa] So the allocation is feasible
——————
2Ra + Rb

Ben Rb

If N is bigger: more blocking coalitions?


The 2-person core
 An arbitrary allocation - can it
be blocked?
 [R]  Draw a line to the endowment
Take N=500 of each tribe
Divide the line for different
 [xb]
coalition numbers


   We’ve found the blocking
 coalition
 If line is not a tangent this
[xa]  can always be done

numbers of…
a-tribe b-tribe

500 360
400
250
450
310

92
04/10/2017

In the limit

[R]

 If N a coalition can be
found that divides the line to
[xb] [R] in any proportion you want

 [x*]  Only if the line is like this will
the allocation be impossible to
 block
[xa]
 With the large N the core has
“shrunk” to the set of CE

Review
 Basic components of trading equilibrium:
• Coalitions
• Blocking
• Core as an equilibrium concept
 Relation to CE
• Every CE must lie in the core
• In the limit of a replication economy the core consists only of CE
 Answer to question: why price-taking?
• In a large economy with suitably small agents…
• …it's the only thing to do

93
04/10/2017

Aggregates
 From household’s demand function  Because incomes depend
xih = Dih(p, yh) on prices
= Dih(p, yh(p) )
 So demands are just functions of p  xih(•) depends on holdings of
xih = xih(p) resources and shares

 If all goods are private (rival) then  “Rival”: extra consumers require
aggregate demands can be written: additional resources. Same as
xi(p) = h xih(p) “consumer: aggregation”

 From firm’s supply of net output  standard supply functions/


demand for inputs
qif = qif(p)
 Aggregate: valid if there are no externalities.
qi = f qif(p) As in “Firm and the market”)

Derivation of xi(p)
 Alf’s demand curve for good 1
 Bill’s demand curve for good 1
 Pick any price
 Sum of consumers’ demand
 Repeat to get the market demand curve

p1 p1 p1

x1a x1b x1
Alf Bill The Market

94
04/10/2017

Derivation of qi(p)
 Supply curve firm 1 (from MC)
 Supply curve firm 2
 Pick any price
 Sum of individual firms’ supply
 Repeat…
 The market supply curve

p p p


q1 q2 q1+q2

low-cost high-cost both firms


firm firm

Subtract q and R from x to get E:

p1 p1 p1
Resource

Demand
stock

Supply
x1 q1 R1

p1

Ei(p) := xi(p) – qi(p) – Ri

E1

95
04/10/2017

Equilibrium in terms of Excess Demand


Equilibrium is characterised by a
price vector p*  0 such that:
 For every good i: The materials balance condition
(dressed up a bit)
Ei(p*) 0

 For each good i that has a If this is violated, then somebody,


positive price in equilibrium somewhere isn't maximising…
(i.e. if pi* > 0):
Ei(p*) = 0 You can only have excess supply
of a good in equilibrium if the
price of that good is 0

Using E to find the equilibrium


 Five steps to the equilibrium allocation
1. From technology compute firms’ net output functions
and profits
2. From property rights compute household incomes and
thus household demands
3. Aggregate the xs and qs and use x, q, R to compute E
4. Find p* as a solution to the system of E functions
5. Plug p* into demand functions and net output functions
to get the allocation
 But this raises some questions about step 4

96
04/10/2017

Two fundamental properties…


 Walras’ Law. For any price p: You only have to work
n with n-1 (rather than n) equations
 pi Ei(p) = 0
i=1

 Homogeneity of degree 0. For any


price p and any t > 0 : You can normalise the prices by
any positive number

Ei(tp) = Ei(p)

Reminder: these hold for any competitive allocation,


not just equilibrium

Price normalisation
 We may need to convert from n numbers p1, p2,…pn to n1
relative prices
 The precise method is essentially arbitrary
 The choice of method depends on the purpose of your model
 It can be done in a variety of ways:

You could divide by


n  This method might seem weird
apnuméraire
pplabour
n
p
MarsBar
i=1
i  But it has a nice property

to give a  The set of all normalised prices


is convex and compact
neat
Mars
set of set
baroftheory
standard
“Marxian”
prices n-1
thatprices
value
theory system
of
ofvalue
sum value
to 1

97
04/10/2017

Normalised prices, n = 2

 The set of normalised prices


p2
 The price vector (0,75, 0.25)

 (0,1)
J={p: p0, p1+p2 = 1}

(0, 0.25)

(0.75, 0)
 p1
(1,0)

The existence problem


 Imagine a rule that moves prices in direction of excess demand:
• “if Ei >0, increase pi”
• “if Ei <0 and pi >0, decrease pi”
• An example of this under “stability” below
 This rule uses the E-functions to map the set of prices into itself
 An equilibrium exists if this map has a “fixed point”
• a p* that is mapped into itself?

 To find the conditions for this, use normalised prices


• pJ
• J is a compact, convex set
• So the mapping has a fixed point

 We can examine this in the special case n = 2


• In this case normalisation implies that p2  1  p1

98
04/10/2017

Existence of equilibrium?
 ED diagram, normalised prices
 Excess demand function with
well-defined equilibrium price
 Case with discontinuous E
 Case where excess demand
for good2 is unbounded below
1
1. E-functions are:
p1
continuous,
bounded below
Excess Excess 2. No equilibrium
supply  p1* demand price where E
crosses the axis
3. E never
0 E1 crosses the
axis

Multiple equilibria

 Three equilibrium prices


 Suppose there were more of resource 1
 Now take some of resource 1 away

1
p1 



E1
0

99
04/10/2017

Adjustment and stability


 Adjust prices according to sign of Ei:
• If Ei > 0 then increase pi
• If Ei < 0 and pi > 0 then decrease pi
 A linear tâtonnement adjustment mechanism:

 Define distance d between p(t) and equilibrium p*


 Given WARP, d falls with t under tâtonnement

Globally stable…
1  Start with a very high price
 Yields excess supply
p1(0)  Under tâtonnement price falls
 Start instead with a low price
 Yields excess demand

Excess Excess  Under tâtonnement price rises

supply  p1* demand


 If E satisfies
p1 WARP then
the system must
converge…
p1(0) E1
0
E1(0) E1(0)

100
04/10/2017

Not globally stable…


 Start with a very high price
1  …now try a (slightly) low price
 Start again with very low price
p1  …now try a (slightly) high price

 Check the “middle” crossing

Excess  Excess
supply demand  Here WARP
does not hold
 Two locally
 stable
equilibria
  One unstable
0
E1

Decentralisation again

x2  The attainable set


 The “Better-than-x* ” set
 The price line
 Decentralisation

 A = {x: x  q+R, (q)  0}

B  B = {hxh: Uh(xh)  Uh(x*h)}


 x*
p1  x* maximises income over A

p2
 x* minimises expenditure
A over B
x1
0

101
04/10/2017

A non-convex technology

 The case with 1 firm

output
 q'
B  Rescaled case of 2 firms,
 … 4 , 8 , 16
 Limit of the averaging process
 The “Better-than” set

• q*  “separating” prices and equilibrium

 Limiting attainable set is


convex

  Equilibrium q* is sustained by
input a mixture of firms at q° and q'

Non-convex preferences
 The case with 1 person
x2  Rescaled case of 2 persons,
 A continuum of consumers
 The attainable set
 “separating” prices and equilibrium

 x'

• x* B
 Limiting better-than set is convex

 Equilibrium x* is sustained by a
mixture of consumers at x° and x'
 x°
A
x1

102
04/10/2017

UNCERTAINTY AND RISK:


Lectures 13 - 15
 Quantities  Functions
x scalar payoff under state  U utility function
x vector payoff under state  Ri u felicity function
resource stock of good i  Other
 state of the world
 Prices and income  set of all states of the world
pi price of good i P prospect
y money income  probability of state 
 certainty-equivalent income  proportionate bond holding
L loss r rate of return
 insurance premium E expectation
 absolute risk aversion
 relative risk aversion

Concepts
 state-of-the-world 
 pay-off (outcome) x X
 prospects {x: }
 ex ante before the realisation
 ex post after the realisation

The time line


The "moment of truth"
The ex-ante view
The ex-post view
time

103
04/10/2017

The state-space diagram: #


The consumption space
xBLUE under uncertainty: 2 states
A prospect in the 1-good 2-state
case
The components of a
prospect in the 2-state case
But this has no equivalent
in choice under certainty

payoff if
RED occurs

 P0

45°
xRED
O

The state-space diagram: #=3


xBLUE

The idea generalises: here we


have 3 states
 = {RED,BLUE,GREEN}

A prospect in the 1-good 3-


state case

•P 0

104
04/10/2017

Another look at preference axioms


 Completeness
to ensure existence
 Transitivity of indifference curves

 Continuity
 Greed
to give shape
 (Strict) Quasi-concavity of indifference curves
 Smoothness

Ranking prospects
xBLUE
Greed: Prospect P1 is preferred to P0
Contours of the preference map

 P1

 P0
xRED
O

105
04/10/2017

Implications of Continuity
xBLUE Pathological preference for certainty (violates
continuity)
Impose continuity
An arbitrary prospect P0
Find point E by continuity
Income  is the certainty equivalent of P0

 E

 P0
xRED
O

Reinterpret quasiconcavity
Take an arbitrary prospect P0
xBLUE
Given continuous indifference curves…
…find the certainty-equivalent prospect E

Points in the interior of the line P0E


represent mixtures of P0 and E
If U strictly quasiconcave P1 is preferred to P0

 E
 P1
 P0
xRED
O

106
04/10/2017

A change in perception
The prospect P0 and certainty-
xBLUE equivalent prospect E (as before)
Suppose RED begins to seem less likely
Now prospect P1 (not P0) appears
equivalent to E
Indifference curves after the change

This alters the slope of the ICs

 E

 
P0 P1
xRED
O

The independence axiom


 Let P(z) and P′(z) be any two distinct prospects such that the
payoff in state-of-the-world  is z
• x = x′ = z

 If U(P(z)) ≥ U(P′(z)) for some z then U(P(z)) ≥ U(P′(z)) for all z

 One and only one state-of-the-world can occur


• So, assume that the payoff in one state is fixed for all prospects
• Level at which payoff is fixed has no bearing on the orderings over
prospects where payoffs differ in other states of the world

 We can see this by partitioning the state space for > 2

107
04/10/2017

Independence axiom: illustration


xBLUE

A case with 3 states-of-the-


world
What if we compare all Compare prospects with the
of these points…? same payoff under GREEN
Ordering of these prospects
should not depend on the size
Or all of these of the payoff under GREEN
points…?

Or all of
these?

The “revealed likelihood” axiom


 Let x and x′ be two payoffs such that x is weakly preferred to x′

 Let 0 and 1 be any two subsets of 

 Define two prospects:


• P0 := {x′ if 0 and x if 0}
• P1 := {x′ if 1 and x if 1}
 If U(P1) ≥ U(P0) for some such x and x′ then U(P1) ≥ U(P0) for
all such x and x′

 Induces a consistent pattern over subsets of states-of-the-world

108
04/10/2017

A key result
 A result that is central to the analysis of uncertainty

 Introducing the three new axioms:


• State irrelevance
• Independence
• Revealed likelihood
 …implies that preferences must be representable in the form of
a von Neumann-Morgenstern utility function:
  ux


 Alternatively, write as Eux


• “expectation” uses the numbers  to weight the payoff evaluations ux

Implications of vNM structure (1)


xBLUE A typical IC
Slope where it crosses the 45º ray?
From the vNM structure
So all ICs have same slope on 45º ray

RED
– _____
BLUE

xRED
O

109
04/10/2017

Implications of vNM structure (2)


xBLUE A given income prospect

From the vNM structure

Mean income

Extend line through P0 and P to P1


 P1


 P
_
 By quasiconcavity U(P)  U(P0)
 P0

xRED
O
Ex

Risk aversion and concavity of u


 Use the interpretation of risk aversion as quasiconcavity
 If individual is risk averse then U( )  U(P0)

 Given the vNM structure…


• u(Ex)  REDu(xRED) + BLUEu(xBLUE)
• u(REDxRED+BLUExBLUE)  REDu(xRED) + BLUEu(xBLUE)

 So the function u is concave

110
04/10/2017

The “felicity” function


Diagram plots utility level
u (u) against payoffs (x)
Payoffs in states BLUE
and RED
u of the average of xBLUE and If u is strictly concave then
xand
REDxRED
equals
higher
the expected
than the u of person is risk averse
BLUE anduof
xexpected ofxxRED
BLUE and of
xRED If u is a straight line then
person is risk-neutral
If u is strictly convex then
person is a risk lover

x
xBLUE xRED

Attitudes to risk
 Shape of u associated
 Risk-neutral with risk attitude
u(x)
 Neutrality: will just accept a
fair gamble
 Aversion: will reject some
better-than-fair gambles
 Loving: will accept some
x unfair gambles
xBLUE xRED
Ex
  u(x)
Risk-averse Risk-loving
u(x)

x x
xBLUE xRED xBLUE xRED
Ex Ex

111
04/10/2017

Risk premium and risk aversion


A given income prospect

xBLUE The certainty equivalent


income
Slope gives probability ratio
Mean income
The risk premium

 Risk premium:
–  Amount that amount you
 P
would sacrifice to eliminate
RED the risk
– _____
 P0 BLUE  Useful additional way of
characterising risk attitude
xRED
O
 Ex

Risk premium: an example


u  Utility values of two payoffs
 Expected payoff and the
utility of expected payoff
 Expected utility and the
certainty-equivalent
u(xRED) u(x)  The risk premium again
u(Ex)
Eu(x )

u(xBLUE)

xRED
x
xBLUE
 Ex

112
04/10/2017

Change the u-function


u
The utility function and risk
premium as before
Now let the utility function
become “flatter”…

u(xRED)

Making the u-function less


curved reduces the risk
u(xBLUE) premium…
…and vice versa
u(xBLUE)
More of this later

x
 
xBLUE xRED
Ex

Absolute and relative risk aversion


 Define absolute and relative risk aversion for scalar payoffs
uxx(x) uxx(x)
(x) :=   ; (x) :=  x 
ux(x) ux(x)
• For risk-averse individuals 
• For risk-neutral individuals 
  independent of scale and origin of u
• can show this from the definitions
•  are two different ways of capturing “curvature” of u
 The definitions are linked:
(x) = x (x)
d(x) d(x)
 = (x) + x 
dx dx

113
04/10/2017

Special cases: CARA and CRRA


1. Constant Absolute Risk Aversion
• Assume that (x) = for all x
• Felicity function must take the form
1
u(x) =   ex

2. Constant Relative Risk Aversion
• Assume that (x) =  for all 
• Felicity function must take the form
1
u(x) =  x1  
1

 Each induces a distinctive pattern of indifference curves…

Constant Absolute Risk Aversion


Case where  = ½
Slope of IC is same along 45° ray (standard vNM)
xBLUE For CARA slope of IC is same along any 45° line

xRED
O

114
04/10/2017

Constant Relative Risk Aversion


Case where  = 2
xBLUE Slope of IC is same along 45° ray (standard vNM)
For CRRA slope of IC is same along any ray
ICs are homothetic

xRED
O

Lotteries
 Consider lottery as a particular type of uncertain prospect
 Take an explicit probability model
 Assume a finite number  of states-of-the-world
 Associated with each state  are:
• A known payoff x ,
• A known probability  ≥ 0
 Lottery is probability distribution over the “prizes” x, =1,2,…,
• The probability distribution is just the vector := (,,…,)
• Of course, + +…+ = 1
 What about preferences?

115
04/10/2017

The probability diagram: #=2


Probability of state RED

BLUE Probability of state BLUE

Cases of perfect certainty


Cases where 0 <  < 1

 (0,1) The case (0.75, 0.25)

 Only points on the purple line


make sense
(0, 0.25) This is an 1-dimensional
• example of a simplex

(0.75, 0)
 (1,0) RED

The probability diagram: #=3


BLUE Third axis corresponds to
probability of state GREEN
There are now three cases of
perfect certainty
 (0,0,1)
Cases where 0 <  < 1
The case (0.5, 0.25, 0.25)

 Only points on the purple


triangle make sense,
(0, 0, 0.25)
•  (0,1,0)
This is a 2-dimensional
(0, 0.25 , 0) example of a simplex

0
(0.5, 0, 0)  (1,0,0)
RED

116
04/10/2017

Preferences over lotteries


 Take probability distributions as objects of choice
• lotteries °, ', ",…
 Each lottery  has same payoff structure
• state-of-the-world  has payoff x
• probability ° or ' or " … depending on which lottery
 Axioms of preference over lotteries 
• Transitivity over lotteries
• If ° ≽ ' and ' ≽ " then °≽"
• Independence of lotteries
• If ° ≽ ' and (0,1) then ° ]" ≽ ' ] "
• Continuity over lotteries
• If ° ≻' ≻ " then there are numbers  and  such that
• °  ]" ≻ ' and ' ≻ °  ]"

Basic result
 Take the axioms transitivity, independence, continuity
 Imply that preferences must be representable in the form of a
von Neumann-Morgenstern utility function:

  ux


 or equivalently:

   


where  ux
 So we can also see the EU model as a weighted sum of s

117
04/10/2017

-indifference curves

 (0,0,1) Indifference curves over probabilities

Effect of an increase in the size of BLUE

 .
 (0,1,0)
(1,0,0)

Trade

118
04/10/2017

Contingent goods: equilibrium trade


b  Certainty line for Alf
xRED
a •
xBLUE
Ob
 Alf's indifference curves
 Certainty line for Bill
 Bill's indifference curves
 Endowment point
 Equilibrium prices & allocation


b
xBLUE
Oa a
xRED

Trade: problems
 Do all these markets exist?
• If there are  states-of-the-world…
• …there are n of contingent goods
• Could be a huge number
 Consider introduction of financial assets
 Take a particularly simple form of asset:
• a “contingent security”
• pays $1 if state  occurs
 Can we use this to simplify the problem?

119
04/10/2017

Attainable set: buying a risky asset


 Endowment
xBLUE  If all resources put into bonds
 All these points belong to A
 Can you sell bonds to others?
 Can you borrow to buy bonds?
 If loan shark willing to finance you

_ _
 P _ _
y y+r′, y+r

_ _
[1+r′ ]y, [1+r]y
_
[1+rº]y  P0

A
xRED
_ _
y [1+r' ]y

Attainable set: insurance


 Endowment
xBLUE  Full insurance at premium 
 All these points belong to A
 Can you overinsure?
 Can you bet on your loss?

_ _
y  P

L–

 P0
y0 – L

A 
xRED
_
y y0

120
04/10/2017

Consumer choice with a variety of financial


assets
 Payoff if all in cash
 Payoff if all in bond 2
 Payoff if all in bond 3, 4, 5,…

xBLUE  Possibilities from mixtures


 Attainable set
 The optimum

 only bonds 4 and 5


1 2 used at the optimum
3

A 4
 P*
5

xRED
7

Problem and its solution

 But corner solutions may also make sense…

121
04/10/2017

Consumer choice: safe and risky assets


Attainable set, portfolio problem
xBLUE  Equilibrium -- playing safe
 Equilibrium - "plunging"
 Equilibrium - mixed portfolio

_ _
y  P P*

P0
A 

xRED
_
y

Results (1)
 Will the agent take a risk?
 Can we rule out playing safe?
 Consider utility in the neighbourhood of  = 0
 Eu( + r) 
———— | = uy( ) E r
 |
 uy is positive
 So, if expected return on bonds is positive, agent will
increase utility by moving away from  = 0

122
04/10/2017

Results (2)
 Take the FOC for an interior solution
 Examine the effect on * of changing a parameter

 For example differentiate E (ruy( + *r)) = 0 w.r.t.


 E (ruyy( + *r)) + E (r2 uyy( + *r)) */ =0
 * – E (ruyy( + *r))
—— = ———————
 E (r2 uyy( + * r))
• Denominator is unambiguously negative
• To sign the numerator we need to impose more structure
• Assume Decreasing ARA
 Theorem: If an individual has a vNM utility function with DARA
and holds a positive amount of the risky asset then the amount
invested in the risky asset will increase as initial wealth increases

An increase in endowment
Attainable set, portfolio problem
xBLUE  DARA Preferences
 Equilibrium
 Increase in endowment
 Locus of constant 
 New equilibrium
_
y+ 
_  P**
y P* o 

xRED
_ _
y y+

123
04/10/2017

A rightward shift in the distribution


 Attainable set, portfolio problem
xBLUE  DARA Preferences
 Equilibrium
 Change in distribution
 Locus of constant 
 New equilibrium

_ _
y  P Po* P**
 

P0
A 

xRED
_
y

An increase in spread
Attainable set, portfolio problem
 Preferences and equilibrium
xBLUE  Increase r′, reduce r

 P* stays put
So  must have reduced
You don’t need DARA for this

_ _ _ _
y+*r′, y+*r
y  P P*

P0
A 

xRED
_
y

124
04/10/2017

WELFARE:
Lectures 16 - 19
 Quantities  Functions
qif net output by f of i  constitution
x ih consumption by h of i Ch cost function of h
R ih ownership by h of resource i Uh utility function of h
Vh indirect utility function of h
 Prices and income vh utility of h as function of 
pi price of good i f production function of firm f
yh money income of h W social welfare function
Th tax revenue raised from h  social evaluation function
 loss  Other
 social state
 set of all social states
utility possibility set

Social objectives
 Two dimensions of social objectives
 Set of feasible social states

 A social preference map?

 Assume we know the


set of all social states
 How can we draw a


social preference map?
Can it be related to
individual preferences?

objective 1

125
04/10/2017

Elements of a constitution
 Social states 
• can incorporate all sorts of information:
• economic allocations,
• political rights, etc
 Individual (extended) preferences over 
•  ≽ ' means that person h thinks state  is at least as good
as state '
 An aggregation rule for the preferences so as to
underpin the constitution
• A function defined on individual (extended) preferences

The social ordering and the constitution


 Where does this ordering come from?
 Presumably from individuals' orderings over 
• assumes that social values are individualistic
 Define a profile of preferences as
• a list of orderings, one for each member of society
• (≽ , ≽ , ≽ , …)
 The constitution is an aggregation function 
• defined on a set of profiles
• yields an ordering ≽
 So the social ordering is ≽ = (≽ , ≽ , ≽ ,…)

126
04/10/2017

Axioms and a result


 Universality
•  should be defined for all profiles of preferences
 Pareto Unanimity
• if all consider that  is better than ', then the social ordering should rank
 as better than '
 Independence of Irrelevant Alternatives
• if two profiles are identical over a subset of  then the derived social
orderings should also be identical over this subset
 Non-Dictatorship
• no one person alone can determine the social ordering

 Kenneth Arrow’s Theorem:


There is no constitution satisfying these axioms

Relaxing universality
 Could it be that the universal domain criterion is just
too demanding?
 Should we insist on coping with any and every set of
preferences, no matter how bizarre?
 Perhaps imposing restrictions on admissible
preferences might avoid the Arrow impossibility result
 However, we run into trouble even with very simple
versions of social states

127
04/10/2017

Alf, Bill, Charlie decide


1-dimensional social states
Scaling of axes is arbitrary
Three possible states
preference

Views about defence spending


Charlie
  Each individual has
 dramatically different views
  But all three sets of
preferences are “single peaked”
How do they decide?
 
 Bill
  Alf Bill Charlie Verdict
Alf
defence ′≻ ?
 spending
′′ ≻ ′?
 ' " ≻ ′′?

Alf, Bill, Charlie decide (2)


 Same states as before
 Same preferences as before
preference

 Now Bill changes his mind


Charlie
  Now one set of preferences is
 Bill no longer “single peaked”
 How do they decide?

 
 Bill
  Alf Bill Charlie Verdict
Alf
defence
′≻ ?
 spending
′′ ≻ ′?
 ' " ≻ ′′?

128
04/10/2017

Alternative voting systems…


 Relaxing IIA involves an approach that modifies the
type of “aggregation rule”
 Simple majority voting may make too little use of
information about individual orderings or preferences
 Here are some alternatives:
• de Borda (weighted voting)
• Single transferable vote
• Elimination voting
 None of these is intrinsically ideal
• Consider the results produced by third example

The IOC Decision Process


 An elimination process
 1997: Appears to give an orderly convergence
• Athens is preferred to Rome irrespective of the presence of other alternatives
 1993: Violates IIA
• Ordering of Sydney, Peking depends on whether other alternatives are
present

129
04/10/2017

A definition of efficiency
 The basis for evaluating social the utility level enjoyed by
states: vh() person h under social state 

 A social state  is Pareto


Note the similarity with the
superior to state ' if: concept of blocking by a
1. For all h: vh()  vh(') coalition
2. For some h: vh() > vh(')

 A social state  is Pareto


efficient if: “feasibility” could be
1. It is feasible determined in terms of the
usual economic criteria
2. No other feasible state
is Pareto superior to 

Derive the utility possibility set


b  From the attainable set…
 …take an allocation
(x1a, x2a)  Evaluate utility for each agent
(x1b, x2b)
 Repeat to get utility possibility set

A A
a=Ua(x1a, x2a)
b=Ub(x1b, x2b)

a

130
04/10/2017

Finding an efficient allocation


 max L( [x ], [q], ) :=
U1(x1) + hh [Uh(x h)  h] f f  f (q f) + i i[qi + Ri  xi]
where xi = h xih, qi = f qi f

 Differentiate L w.r.t xih. If xih, xjh positive at the optimum:


hUih(xh) = i hUjh(xh) = j

 Differentiate L w.r.t qif. If qif , qjf nonzero at the optimum:


f if (qf) = i f jf (qf) = j

Interpreting the FOC


 From the FOCs for any household
h and goods i and j:

Uih(xh) i for every household:


———— = — MRS = shadow price ratio
Ujh (xh) j

 From the FOCs for any firm f


and goods i and j:

if(qf) i for every firm:


———— = — MRT = shadow price ratio
jf (qf) j

131
04/10/2017

Efficiency in an Exchange Economy


x1b
Ob
x2a  Alf’s indifference curves
 Bill’s indifference curves
 The contract curve

Set of efficient
allocations is the
Allocations where contract curve
MRS12a = MRS12b
Includes cases where
Alf or Bill is very poor

x2b
Oa
x1a

Efficiency with production

q2f x2h  Household h’s indifference curves


 h’s consumption in the efficient allocation
 MRS
 Firm f’s technology set
 f’s net output in the efficient allocation

MRS = MRT at efficient point


^ fh
x
q

x1h
0 q1f

132
04/10/2017

Two welfare theorems


 Welfare theorem 1
• Assume a competitive equilibrium
• What is its efficiency property?
 THEOREM: if all consumers are greedy and there are no
externalities then a competitive equilibrium is efficient

 Welfare theorem 2
• Pick any Pareto-efficient allocation
• Can we find a property distribution d so that this allocation is a CE for d?
 THEOREM: if, in addition to conditions for theorem 1, there are
no non-convexities then an arbitrary PE allocation be supported
by a competitive equilibrium

Supporting a PE allocation
x1b

Ob
x2a [R]  The contract curve
 An efficient allocation
 Supporting price ratio = MRS
 The property distribution
 A lump-sum transfer

Allocations where ^ p1
 [x] 
MRS12a = MRS12b
p2  Support allocation by a CE
 This needs adjustment of
the initial endowment
 Lump-sum transfers may be
tricky to implement

x2b
Oa
x1a

133
04/10/2017

Individual household behaviour

x2h  Household h’s indifference curves


 h’s consumption in the efficient
allocation
 Supporting price ratio = MRS

 h’s consumption in the allocation


is utility-maximising for h

^xh
 h’s consumption in the allocation
p1 is cost-minimising for h

p2

x1h
0
March 2012

Supporting a PE allocation (production)


q2f  Firm f’s technology set
 f’s net output in the efficient allocation
 Supporting price ratio = MRT

 f’s net output in the allocation is


^f
q profit-maximising for f
p1

p2

q1f
0

134
04/10/2017

Firm f makes “wrong” choice


 Firm f ’s production function
~f
q2f  q violates second theorem
 Suppose we want to allocate this
net output to f
 Introduce prices
 f's choice at those prices

 A twist on the previous


^f
q example
p1
  Big fixed-cost component to
p2 producing good 1

 “market failure” once again


q1f
0
March 2012

PE allocations – two issues


good 2


 Same production function
 Is PE here…?
 or here?

 Implicit prices for MRS=MRT

 Competitive outcome


 Issue 1 – what characterises
the PE?
 Issue 2 – how to implement
the PE
good 1
0
March 2012

135
04/10/2017

Indecisiveness of PE
b
 Construct utility-possibility set
as previously
 Two efficient points
 Points superior to 
 Points superior to '

Boundary points cannot be


compared on efficiency grounds
v
   and ' cannot be compared
on efficiency grounds
 v
a

“Potential” Pareto superiority


 Define  to be potentially superior to ' if :
• there is a * which is actually Pareto superior to '
• * is “accessible” from 
 To make use of this concept we need to define accessibility
• use a tool from the theory of consumer welfare
 CVh('  ): the monetary value the welfare change for person h…
• …of a change from state ' to state 
• …valued in terms of the prices at 
 CVh > 0 means a welfare gain; CVh < 0 a welfare loss

 THEOREM: a necessary and sufficient condition for  to be


potentially superior to ' is
h CVh('  ) > 0

136
04/10/2017

Applying potential superiority


b   is not superior to ' and ' is
not superior to 
 * is superior to 
 Points accessible from '

v*  There could be lots of points


 accessible by lump-sum transfers
v  ' is potentially superior to 

 v
a

Re-examine potential superiority


 The process from  to ', as before

b  The process in reverse from ' to 


points accessible  Combine the two
from 

 is potentially superior to  and …


v  is potentially superior to !

points accessible
from 
 v

a

137
04/10/2017

Price distortion: efficiency loss


Production possibilities
x2 An efficient allocation
Some other inefficient allocation

 At x* producers and consumers


face same prices
•x  At x producers and consumers
face different prices
•x*  Price "wedge" forced by the
distortion

p*

x1
0

Waste measurement: a method


 To measure loss we use a reference point
 Take this as competitive equilibrium…
• …which defines a set of reference prices
 Quantify the effect of a notional price change:
• pi := pi – pi*
• This is [actual price of i] – [reference price of i]
 Evaluate the equivalent variation for household h :
• EVh = Ch(p*,h) – Ch(p, h) – [y*h – yh]
• This is (consumer costs) – (income)
 Aggregate over agents to get a measure of loss, 

138
04/10/2017

If producer prices constant…


Production possibilities
C(p, ) x2
Reference allocation and prices

Actual allocation and prices
Cost of  at prices p

Cost of  at prices p*
Change in valuation of output


 Measure cost in terms of good 2
C(p*, ) •x
 Losses to consumers are
C(p*, )  C(p, )
•x*   is difference between
|C(p*, )  C(p, )| and 
p
p*

0 x1

A model of a commodity tax


p1 Equilibrium price and quantity
compensated The tax raises consumer price…
demand curve …and reduces demand
 Gain to the government
 Loss to the consumer
 Waste

Waste given by size of triangle


p1 Sum over h to get total waste
Known as deadweight loss of tax
p1*

x1 *
x1 h
x1 h

139
04/10/2017

Tax: computation of waste


 The tax imposed on good 1 forces a price wedge
• p1 = tp1* > 0 where p1* is the untaxed price of the good
 h’s demand for good 1 is lower with the tax:
• x1** = x1* x1h and x1h < 0
 Revenue raised by government from h:
• Th = tp1* x1**= x1**p1 > 0
 Absolute size of loss to h is
• h= ∫ x1h dp1 ≈ x1** p1 − ½ x1hp1 = Th − ½ t p1* x1h > Th
 Use the definition of elasticity
•  := p1x1h / x1hp1< 0
 Net loss from tax (for h) is
• h = h− Th = − ½tp1* x1h = − ½tp1x1** = − ½t Th
 Overall net loss from tax (for h) is ½ |tT

Size of waste depends upon elasticity


p1
p1 compensated
Redraw previous example

demand curve  low: relatively small waste


 high: relatively large waste

p1
p1*

x1h

p1 x1h
p1
p1

p1 *

p1 p1
p1*
p1 *
x1 h
x1h
x1h x1h
x1h x1h

140
04/10/2017

Fairness in the trading model


x1b
Ob
[x°°] 
x2a [x′′] 
 The Edgeworth box
 Extreme, efficient allocations
 Two more efficient allocations
 Another, intermediate example

  Swap a's and b's allocations

 [x]
 Are [x°], [x°°] "obviously"
unfair?
 Perhaps also [x'], [x''] ?
 a prefers to have b's
allocation in [x]

 [x′]
 So [x] is not fair

[x°] x2b
O a
x1a

Towards a definition of fairness


 Recall the definition of Pareto superiority as:
• allocation [x] is superior to [x′] if…
• for all h: Uh(xh)  Uh(x′h)
• for some h: Uh(xh)  Uh(x′h)
 Use this individualistic approach to formalise fairness as “no-
envy”
• compare, not with an alternative, hypothetical bundle…
• ..but with the bundles enjoyed by other people
 An allocation is fair if, for every pair of individuals h and k:
• Uh(xh)  Uh(xk )
• given my tastes I weakly prefer my bundle to yours

141
04/10/2017

A result on fairness
 THEOREM: if all persons have equal incomes then a competitive
equilibrium is a fair allocation
• An apparently appealing result
• Seems to combines two opposing principles:
• individualism – embodied in competitive behaviour
• egalitarianism – embodied in equal-incomes requirement

 Proof:
• For every household h let Ah := {xh: i pixih  yh }
• If [x*] is a CE then x*h  Ah and Uh(x*h) Uh(xh ) for all xh  Ah
• But if all incomes are equal then, for any h and k: Ah = Ak, so x*k  Ah
• Therefore Uh(x*h) Uh(x*k ) for any households h and k
• So no one would prefer another person’s bundle. CE is fair (envy free)

The fair allocation


x1b
Ob
x2a
 The Edgeworth box
 An efficient allocation
 Supporting price ratio = MRS
 Incomes in terms of good 1

 Allocation [x*] is CE if
 [x*] incomes are as shown

x2b
Oa
x1a

142
04/10/2017

Using a SWF
b
 Take the utility-possibility set
 Social welfare contours
 A social-welfare optimum?
W(a, b,... )

 W defined on utility levels


Not on orderings

• Imposes several restrictions…


..and raises several questions
a

“Veil of ignorance”: formalisation


 Individualistic welfare: use theory of choice under
uncertainty to find shape of W
W(1, 2 , 3 , ...)
 vN-M form of utility function:
probability assigned to 
 u(x)
u : cardinal utility function,
Equivalently:
independent of 
  utility payoff in state 
 Replace  by set of identities
{1,2,..., nh}: welfare is expected utility
h h h from a "lottery on identity“

 A suitable assumption about


“probabilities”? An additive form of the
nh welfare function
1
W = —  h
nh h=1

143
04/10/2017

Questions about “equal ignorance”


 Construct a lottery on identity
 The “equal ignorance” assumption
 Where people know their identity with
certainty
h
 Intermediate case

The “equal ignorance”


assumption: h = 1/nh
But is this appropriate?

Or should we assume that


| | | | | people know their identities
1 2 3 identity h nh with certainty?

Or is the "truth" somewhere


between?

From an allocation to social welfare


(x1a, x2a)
(x1b, x2b)
 From the attainable set...

a=Ua(x1a, x2a)  ...take an allocation


b=Ub(x1b, x2b)  Evaluate utility for each agent
A A
 Plug into W to get social welfare

W(a, b)

 Take the individualistic welfare model


W(1, 2, 3, ...)
 Assume everyone is selfish:
h = Uh(xh) , h=1,2,..., nh
 Substitute in the above:
W(U1(x1), U2(x2), U3(x3), ...)
 What happens to welfare if we vary the allocation in A?

144
04/10/2017

Varying the allocation


 Differentiate w.r.t. xih : The effect on h if
dh = Uih(xh) dxih commodity i is changed

 Sum over i:
n The effect on h if all
dh =  Uih(xh) dxih commodities are changed
i=1

 Differentiate W with respect to h: Changes in utility


nh
change social welfare
dW = Wh dh
h=1

 Substitute for dh in the above: So changes in allocation


nh n
change welfare
dW =  Wh Uih(xh) dxih
h=1 i=1

The SWF maximum problem


 First component of the problem: The objective function
W(U1(x1), U2(x2), U3(x3), ...)

 Second component of the problem: Feasibility constraint


n
(x)  0, xi =  h=1 xihh

 The Social-welfare Lagrangian: n


Constraint subsumes
W(U1(x1), U2(x2),...) −  (h=1xh )
h technological feasibility and
materials balance

 FOCs for an interior maximum: From differentiating Lagrangian


Wh (...) Uih(xh) − i(x) = 0 with respect to xih

145
04/10/2017

Solution to SWF maximum problem


 From FOCs:
MRS equated across all h
Uih(xh) Uiℓ(xℓ)
——— = ——— We’ve met this condition
Ujh(xh) Ujℓ(xℓ) before - Pareto efficiency

 Also from the FOCs: social marginal utility of ice


cream equated across all h
Wh Uih(xh) = Wℓ Uiℓ(xℓ)

 Relate marginal utility to prices: This is valid if all consumers


Uih(xh) = Vyhpi optimise

 Substituting into the above: At optimum the welfare value of


Wh Vyh = Wℓ Vyℓ $1 is equated across all h. Call
this common value M

Social welfare, income, expenditure


 Social welfare can be expressed as: SWF in terms of direct utility
W(U1(x1), U2(x2),...) = W(V1(p,y1), V2(p,y2),...) and in terms of indirect utility

 Differentiate the SWF w.r.t. {yh}: Changes in utility and change


nh nh nh social welfare related to income
dW =  Wh dh= WhVyh dyh= M  dyh
h=1 h=1 h=1

 Differentiate the SWF w.r.t. pi : Changes in utility and change


nh nh social welfare related to prices
WhVihdpi = – WhVyh xihdpi = – M xihdpi
nh
dW =
h=1 h=1 h=1

 THEOREM: in the neighbourhood of a


welfare optimum welfare changes are
measured by changes in national income /
national expenditure .

146
04/10/2017

Income-distribution space: nh=2

 The income space: 2 persons


income
Bill's

An income distribution

 Note the similarity with


diagrams used in the analysis
of uncertainty

y

45° Alf's
O Alf's
income
income

Welfare contours
 An arbitrary income distribution
yb  Contours of W
 Swap identities
 Distributions with the same mean
 Equally-distributed-equivalent income

 Anonymity implies symmetry of W



 Ey is mean income
 Richer-to-poorer income
Ey transfers increase welfare
   is income that, if received
y uniformly by all, would yield same
level of social welfare as y
 Ey  is income that society would
ya
give up to eliminate inequality
 Ey

147
04/10/2017

A result on inequality aversion


 Principle of Transfers : “a mean-preserving redistribution from
richer to poorer should increase social welfare”

 THEOREM: Quasi-concavity of W implies that social welfare


respects the “Transfer Principle”

Special form of the SWF


 It can make sense to write W in the additive form
nh
1
W=—  yh
n h=1 h

• where the function  is the social evaluation function


• (the 1/nh term is unnecessary – arbitrary normalisation)
• Counterpart of u-function in choice under uncertainty
 Can be expressed equivalently as an expectation:
W = E yh
• where the expectation is over all identities
• probability of identity h is the same, 1/nh , for all h
 Constant relative-inequality aversion:
1 1–
y = ——
1–
y
• where  is the index of inequality aversion
• works just like , the index of relative risk aversion

148
04/10/2017

Social views: inequality aversion


 Mild inequality aversion
yb yb
 Strong inequality aversion
 ½
 Priority to poorest
 Indifference to inequality

 Atkinson SWF, general


case ( ):
O ya O ya nh
yb yb W = [yh]1  / [1  ]
h=1
   “Rawlsian” case ( ):
W = min yh
h

 “Benthamite” case ( = 0):


nh
O ya O ya W=  yh
h=1

Social values and welfare optimum


yb  The income-possibility set Y
 Welfare contours (  = 0)
 Welfare contours (  = ½)
 Welfare contours (  = )

Y derived from set A


Nonconvexity, asymmetry come
from heterogeneity of households

 y* maximises total income


Y y*** irrespective of distribution

 y** trades off some income for
**
y greater equality

y*  y*** gives priority to equality; then


 ya maximises income subject to that

149

Вам также может понравиться