Managerial Economics

The Nature and Scope
of Managerial
Economics
Thomas F. Rutherford
1.1
The Course Lectures Hirschey
Lecture 1
The Nature and Scope of Managerial
Economics
Getting Started with Economics
Managerial Economics
September 23, 2010
Center for Energy Policy and Economics
Department of Management, Technology and Economics
ETH Zrich
of Managerial
Economics
1.2
Introduction to Ideas
Let us begin with the ideas of economics in general and then link in
managerial economics.
Economics is: the study of how people allocate scarce
resources.
Managerial economics focuses on how managers allocate their
scarce resources:
People
Skills
Ofce equipment
Warehouses
Machinery
Raw materials
of Managerial
Economics
1.2
resources.
scarce resources:
People
Skills
Ofce equipment
Warehouses
Machinery
Raw materials
of Managerial
Economics
1.2
resources.
scarce resources:
People
Skills
Ofce equipment
Warehouses
Machinery
Raw materials
of Managerial
Economics
1.2
resources.
scarce resources:
People
Skills
Ofce equipment
Warehouses
Machinery
Raw materials
of Managerial
Economics
1.3
Subjects in Business Administration
Courses dealing with the functions of a business:
Production (operations)
Human resources management
Marketing
Finance
There are separate courses for each of these areas, but managerial
economics is not in this list.
of Managerial
Economics
1.3
Subjects in Business Administration
Courses dealing with the functions of a business:
Production (operations)
Human resources management
Marketing
Finance
of Managerial
Economics
1.4
Courses Dealing with the Business Environment
Ethics
Legal issues
International business
Information technology
of Managerial
Economics
1.4
Courses Dealing with the Business Environment
Ethics
Legal issues
International business
Information technology
of Managerial
Economics
1.5
Courses Dealing with Methodology
Quantitative methods
Decision theory and management science
Game theory
Managerial economics
of Managerial
Economics
1.5
Courses Dealing with Methodology
Quantitative methods
Decision theory and management science
Game theory
Managerial economics
of Managerial
Economics
1.6
What is the idea of managerial economics
Use economic princples to solve the problems which managers
encounter when running their businesses:
Tend to be more technical
Involves more mathematics and statistics than other courses.
of Managerial
Economics
1.7
Scope of Managerial Economics
What are the elements of managerial economics and how are these
inter-related?
The core focus of ME is pricing (price theory),
But this view can be misleading (too narrow) ME is generally
concerned with all aspects of rm operation which affect prot.
of Managerial
Economics
1.8
A Taxonomy
Joint dependence of demand and supply:
PRICING
/ \
DEMAND SUPPLY
Basic ideas here:
Supply is cost theory
Demand is theory of the consumer.
of Managerial
Economics
1.9
Other Ideas
Supply is also affected by behavior of producers including
competing rms (production theory and strategy).
Theory of the rm (nature, objectives) interrelates game theory
and business strategy.
Theory of markets concerns the nature of competition (how are
prices and prots determined in different types of competitive
situations?
of Managerial
Economics
1.10
Diagrammatic Perspective
Government
|
----------------------------------------
| |
Theory --------- Pricing ------------ Competitionn
Firm / \ Theory
/ \
Demand Supply
| |
Consumer Production
Theory Theory
of Managerial
Economics
1.11
Lecture Sequence
1 Introductory concepts: scope and context, demand, supply and
market equilibrium.
2 Demand theory and estimation (marketing)
3 Cost and market structure (strategic decisions)
4 Decision making with risk (investment under uncertainty)
of Managerial
Economics
1.12
Outline of Lectures
9/23 The nature and scope of managerial economics (MH 1,2)
9/30 Demand, supply and market equilibrium (MH 3, HW #1)
10/7 Budget constraints, preferences and consumer choice (MH 4)
10/14 Demand functions price and income elasticities (HW # 2)
10/21 Demand estimation and forecasting (MH 5 and 6)
10/28 Case study: marketing (Professor Hoffman)
11/4 Firm level cost minimization (MH: 7 and 8)
11/11 Competitive markets (MH: 10 and 11; HW # 3)
11/18 Case study: rm-level decisions (Professor Hoffman)
11/25 Imperfectly competitive markets (MH 12 and 13)
12/2 Game theory and pricing (MH 14 and 15, HW # 4)
12/9 Risk and uncertainty (MH 16)
12/16 Case study: investment under uncertainty (Professor
Hoffman)
12/23 Final Examination Review
of Managerial
Economics
1.13
Assessment
Homeworks (40%)
Final examination in early January, 2011 (60%).
of Managerial
Economics
1.14
Other information
Instructors: Thomas F. Rutherford and Volker Hoffman
Email: trutherford@ethz.ch
Ofce: ZUE (E7)
Phone: 044 632 6359
Ofce Hours: Wednesday mornings and by appointment.
Course Web Page:
http://ethz.ch/cepe/education/managerialeconomics
of Managerial
Economics
1.15
Textbook
The course will be based on the textbook Managerial Economics
(12th edition) by Mark Hirschey (MH).
Additional readings will be periodically assigned from Managerial
Economics: A Problem-Solving Approach (2nd edition) by Froeb
and McCann, South-Western.
Copies of these texts are available for short-term loan from my
secretary Rina Fichtl, ZUE E8 (rfichtl@ethz.ch).
of Managerial
Economics
1.16
Objectives of the Course
Learn foundations of economics
Appreciate the role of economic ideas in managerial decisions.
Learn some formal models and methods of analysis in
economics and management science.
of Managerial
Economics
1.17
Key Ideas from Herschey Chapter 1
How Is Managerial Economics Useful?
Theory of the Firm
Prot Measurement
Why Do Prots Vary among Firms?
Role of Business in Society
of Managerial
Economics
1.18
Key Concepts
managerial economics
theory of the firm
expected value maximization
value of the firm
present value
optimize
satisfice
business profit
normal rate of return
economic profit
profit margin
return on stockholders equity
frictional profit theory
monopoly profit theory
innovation profit theory
compensatory profit theory
of Managerial
Economics
1.19
How is Managerial Economics Useful?
Evaluating Choice Alternatives
Identify ways to efciently achieve goals.
Specify pricing and production strategies.
Spell out production and marketing rules to maximize prots.
Making the Best Decision
Managerial economics helps meet management objectives
efciently.
Managerial economics shows the logic of consumer, and
government decisions
of Managerial
Economics
1.20
of Managerial
Economics
1.21
Theory of the Firm
Expected Value Maximization
Owner-managers maximize short-run prots.
Primary goal is long-term expected value maximization.
Constraints and the Theory of the Firm
Resource constraints.
Social constraints.
Limitations of the Theory of the Firm
Alternative theory adds perspective.
Competition forces efciency.
Hostile takeovers threaten inefcient managers.
of Managerial
Economics
1.22
Measuring Prot
Business Versus Economic Prot
Business (accounting) prot reects explicit costs and revenues.
Economic prot.
Prot above a risk-adjusted normal return.
Considers cash and noncash items.
Variability of Business Prots
Business prots vary widely
of Managerial
Economics
1.23
Why Do Prots Vary Among Firms?
Disequilibrium Prot Theories
Unexpected revenue growth.
Unexpected cost savings.
Compensatory Prot Theories
Prots accrue to rms that are better, faster, or cheaper than the
competition.
of Managerial
Economics
1.24
Role of Business in Society
Why Firms Exist
Businesses help satisfy consumer wants.
Businesses contributes to social welfare
Social Responsibility of Business
Serve customers.
Provide employment opportunities.
Play by the rules (laws and regulations)
of Managerial
Economics
1.25
Froeb and McCann:
Managerial Economics and Problem Solving
Problemsolvingrequirestwosteps:First,figureoutwhymistakesare
beingmade;andthenfigureouthowtomakethemstop.
Therationalactorparadigmassumesthatpeopleactrationally,optimally,
andselfinterestedly.Tochangebehavior,youhavetochangeincentives.
Goodincentivesarecreatedbyrewardinggoodperformance.
Awelldesignedorganizationisoneinwhichemployeeincentivesare
alignedwithorganizationalgoals.Bythiswemeanthatemployeeshave
enoughinformationtomakegooddecisions,andtheincentivetodoso.
Youcananalyzeanyproblembyaskingthreequestions:(1)Whoismaking
thebaddecision?;(2)Doesthedecisionmakerhaveenoughinformation
tomakeagooddecision?;and(3)theincentivetodoso?
Answerstothesequestionswillsuggestsolutionscenteredon(1)letting
someoneelsemakethedecision,someonewithbetterinformationor
incentives;(2)givingthedecisionmakermoreinformation;or(3)changing
thedecisionmakersincentives.
Problem:OverbiddingOVIgastract
Ayounggeologistwaspreparingabidrecommendationforanoiltractin
theGulfofMexico.
Withknowledgeoftheproductivityofneighboringtractsalsoownedby
company,thegeologistrecommendedabidof$5million.
Seniormanagement,though,bid$20million faroverthenexthighest
bidof$750,000.
What,ifanything,iswrong?
Thegoalofthistextistoprovidetoolstohelpdiagnoseandsolve
problemslikethis.
2
Problemsolving
Twodistinctsteps:
Figureoutwhatswrong,i.e.,whythebaddecisionwas
made
Figureouthowtofixit
Bothstepsrequireamodelofbehavior
Whyarepeoplemakingmistakes?
Whatcanwedotomakethemchange?
Economistsusetherationalactorparadigmtomodel
behavior.Therationalactorparadigmstates:
Peopleactrationally,optimally,selfinterestedly
i.e.,theyrespondtoincentives tochangebehavioryoumust
changeincentives.
3
Howtofigureoutwhatiswrong
Undertherationalactorparadigm,mistakesaremadeforone
oftworeasons:
lackofinformationor
badincentives.
Todiagnoseaproblem,ask3questions:
1.Whoismakingbaddecision?
2.Dotheyhaveenoughinfotomakeagooddecision?
3.Dotheyhavetheincentivetodoso?
4
Howtofixit
Theanswerswillsuggestoneormoresolutions:
1.Letsomeoneelsemakethedecision,someonewithbetter
informationorincentives.
2.Changetheinformationflow.
3.Changeincentives
Changeperformanceevaluationmetric
Changerewardscheme
Usebenefitcostanalysistochoosethebest(most
profitable?)solution
5
Keeptheultimategoalinmind
Forabusinessororganizationtooperateprofitablyand
efficientlytheincentivesofindividualsneedtobealigned
withthegoalsofthecompany.
Howdowemakesureemployeeshavetheinformation
necessarytomakegooddecisions?
Andtheincentive todoso?
6
Analyzetheoverbiddingmistake
Anotherclue:
Afterwinningthebid,thegeologistincreasedtheestimated
reservesofthecompany.
But,afteradrywellwasdrilled,thereserveestimateswere
decreased.
SeniorManagementsteppedinandorderedanincreaseinthe
reserveestimate.
Lastclue:
Seniormanagementresignedseveralmonthslater.
7
ANSWER:Managerbonusesfor
increasingreserves
Thebonussystemcreatedincentivestooverbid.
Seniormanagerswererewardedforacquiringreserves
regardlessoftheirprofitability
Bonusesalsocreatedincentivetomanipulatethe
reserveestimate.
Nowthatweknowwhatiswrong,howdowefixit?
Letsomeoneelsedecide?
Changeinformationflow?
Changeincentives?
Performanceevaluationmetric
Rewardscheme
8
Ethics
Doestherationalactorparadigmencourageselfinterested,
selfishbehavior?
NO!
Opportunisticbehaviorisafactoflife.
Youneedtounderstanditinordertocontrolit.
Therationalactorparadigmisatoolforanalyzingbehavior,not
aprescriptionforhowtoliveyourlife.
9
Economic Analysis of
Competitive Markets
2.4
Lecture Overview Microeconomics Review Examples of applied price theory Building a Microeconomic Model Demand and Supply Equilibrium Taxation Elasticities Consumer and Producer Surplus
Market equilibrium
A market is in equilibrium when total quantity demanded by
buyers equals total quantity supplied by sellers.
An equilibrium is supported by market prices.
At equilibrium prices, the market is made up of voluntary
participants.
Market prices reect marginal willingness to accept (by rms)
and marginal willingness to pay (by consumers).
Competitive Markets
2.5
Willingness to Pay (=Marginal Value)
p
D(p)
Market Demand
q = D(p)
Competitive Markets
2.6
Willingness to Accept (=Marginal Cost)
p
S(p)
Market Supply
q = S(p)
Competitive Markets
2.7
Equilibrium
p
S(p), D(p)
Market Supply
q = S(p)
Market Demand
q = D(p)
p
Competitive Markets
2.8
Disequilibrium Price Above Equilibrium: Excess Supply
p
S(p), D(p)
Market Supply
q = S(p)
Market Demand
q = D(p)
p
D(p
) S(p
)
p
D(p
) < S(p
):
Excess supply
Competitive Markets
2.9
Disequilibrium Price Below Equilibrium: Excess Demand
p
S(p), D(p)
Market Supply
q = S(p)
Market Demand
q = D(p)
p
D(p
) S(p
)
p
D(p
) > S(p
):
Excess demand
Competitive Markets
2.10
Equilibrium in a Linear Model
An example of calculating a market equilibrium when the market
demand and supply curves are linear:
D(p) = a bp
S(p) = c +dp
Hence:
a bp
c +dp
and the equilibrium price is:

p
=
a c
b +d
and the equilibrum quantity is:
q
= D(p
) = S(p
) =
ad +bc
b +d
Competitive Markets
2.10
D(p) = a bp
S(p) = c +dp
Hence:
a bp
c +dp

p
=
a c
b +d
q
= D(p
) = S(p
) =
ad +bc
b +d
Competitive Markets
2.10
D(p) = a bp
S(p) = c +dp
Hence:
a bp
c +dp

p
=
a c
b +d
q
= D(p
) = S(p
) =
ad +bc
b +d
Competitive Markets
2.10
D(p) = a bp
S(p) = c +dp
Hence:
a bp
c +dp

p
=
a c
b +d
q
= D(p
) = S(p
) =
ad +bc
b +d
Competitive Markets
2.11
Inverse Demand and Supply
Willingness to pay can be characterized by representing price as an
inverse function of quantity:
q = D(p) = a bp p =
a q
b
= D
1
(q)
and willingness to accept is likewise dened:
q = S(p) = c +dp p =
c +q
d
= S
1
(q)
Competitive Markets
2.12
D
1
(q)
S
1
(q)
S(p), D(p)
Market Inverse Supply
S
1
(q) = (c +q)/d
D
1
(q) = (a q)/b
Market
Inverse
Demand
p
Competitive Markets
2.13
Dual Formulation
In equilibrium, we have rms supply to the point where market price
equals willingness to supply:
p = S
1
(q) =
c +q
d
and households consume goods to the point where market price
equals willingness to pay:
p = D
1
(q) =
a q
b
= S
1
(q) =
c +q
d
Hence, in equilibrium
S
1
(q) = D
1
(q)
and
q
=
ad +bc
b +d
so
p
= D
1
(q
) = S
1
(q
) =
a c
b +d
Competitive Markets
2.13
Dual Formulation
p = S
1
(q) =
c +q
d
p = D
1
(q) =
a q
b
= S
1
(q) =
c +q
d
S
1
(q) = D
1
(q)
and
q
=
ad +bc
b +d
so
p
= D
1
(q
) = S
1
(q
) =
a c
b +d
Competitive Markets
2.13
Dual Formulation
p = S
1
(q) =
c +q
d
p = D
1
(q) =
a q
b
= S
1
(q) =
c +q
d
S
1
(q) = D
1
(q)
and
q
=
ad +bc
b +d
so
p
= D
1
(q
) = S
1
(q
) =
a c
b +d
Competitive Markets
2.13
Dual Formulation
p = S
1
(q) =
c +q
d
p = D
1
(q) =
a q
b
= S
1
(q) =
c +q
d
S
1
(q) = D
1
(q)
and
q
=
ad +bc
b +d
so
p
= D
1
(q
) = S
1
(q
) =
a c
b +d
Competitive Markets
2.13
Dual Formulation
p = S
1
(q) =
c +q
d
p = D
1
(q) =
a q
b
= S
1
(q) =
c +q
d
S
1
(q) = D
1
(q)
and
q
=
ad +bc
b +d
so
p
= D
1
(q
) = S
1
(q
) =
a c
b +d
Competitive Markets
2.14
Special Case 1: Fixed Supply Quantity
Hal Varian, Intermediate Microeconomics Norton
Competitive Markets
2.15
Fixed Supply Equilibrium
Supply is xed (q
= c), hence price is determined by the inverse

demand curve:
p
= D
1
(q
) =
a c
b +d
Notice that this equilibrium outcome describes a situtation in which
rms are unable to respond to changes in market price, as is quite
common in short-run situations particularly for energy markets in
which changes to infrastructure require many years.
Competitive Markets
2.16
Special Case 2: Fixed Supply Price
Competitive Markets
2.17
Quantity Taxes
A quantity tax levied at a rate of t is a tax of t CHF paid on each
unit traded.
If the tax is levied at on sellers then it is an excise tax.
If the tax is levied on buyers then it is a sales tax.
When a tax is denominated in currency units, it is a specic tax.
When it is denominted as a percentage of the sales value, it is
referred to as an ad-valorem tax.
Competitive Markets
2.18
Quantity Taxes
Typical questions which arise concerning quantity taxes:
What is the effect of a quantity tax on a markets equilibrium?
How are prices affected?
How is the quantity traded affected?
Who pays the tax?
How are gains-to-trade altered?
Competitive Markets
2.19
Market Equilibrium with Quantity Taxes
A tax rate t makes the price paid by buyers, p
b
, higher by t than the
price received by sellers, p
s
:
p
b
= p
s
t
Even with a tax, market clear.
I.e. quantity demanded by buyers at price p
b
must equal quantity
supplied by sellers at price p
s
:
D(p
b
) = S(p
s
)
The market equilibrium then involves two equations in two unknowns.
Notice that these two conditions apply regardless of whether the tax
is levied on sellers or on buyers. Hence, a sales tax rate $t has the
same effect as an excise tax rate $t.
Competitive Markets
2.19
b
s
:
p
b
= p
s
t
b
must equal quantity
s
:
D(p
b
) = S(p
s
)
Competitive Markets
2.19
b
s
:
p
b
= p
s
t
b
must equal quantity
s
:
D(p
b
) = S(p
s
)
Competitive Markets
2.19
b
s
:
p
b
= p
s
t
b
must equal quantity
s
:
D(p
b
) = S(p
s
)
Competitive Markets
2.20
Geometry of Taxation
Competitive Markets
2.21
Geometry of Taxation
Competitive Markets
2.22
Equivalent Impacts of Sales and Excise Taxes
Competitive Markets
2.23
Tax Incidence
Who pays the tax of $t per unit traded?
The division of the $t between buyers and sellers is the incidence
of the tax.
Competitive Markets
2.24
Algebra of Tax Incidence
Equilibrium conditions:
p
b
= p
s
+t
a bp
b
= c +dp
s
Substitute for p
b
in the second equation:
a b(p
s
+t ) = c +dp
s
p
s
=
a c bt
b +d
.
Substitute into the demand or supply function to obtain:
q
t
=
ad +bc bdt
b +d
and
p
b
= p
s
+t =
a c +dt
b +d
Note that as t 0, p
b
p
, the equilibrium price without taxes, and

q
t

ad+bc
b+d
Competitive Markets
2.24
p
b
= p
s
+t
a bp
b
= c +dp
s
Substitute for p
b
a b(p
s
+t ) = c +dp
s
p
s
=
a c bt
b +d
.
q
t
=
ad +bc bdt
b +d
and
p
b
= p
s
+t =
a c +dt
b +d
Note that as t 0, p
b
p

q
t

ad+bc
b+d
Competitive Markets
2.24
p
b
= p
s
+t
a bp
b
= c +dp
s
Substitute for p
b
a b(p
s
+t ) = c +dp
s
p
s
=
a c bt
b +d
.
q
t
=
ad +bc bdt
b +d
and
p
b
= p
s
+t =
a c +dt
b +d
Note that as t 0, p
b
p

q
t

ad+bc
b+d
Competitive Markets
2.24
p
b
= p
s
+t
a bp
b
= c +dp
s
Substitute for p
b
a b(p
s
+t ) = c +dp
s
p
s
=
a c bt
b +d
.
q
t
=
ad +bc bdt
b +d
and
p
b
= p
s
+t =
a c +dt
b +d
Note that as t 0, p
b
p

q
t

ad+bc
b+d
Competitive Markets
2.25
Comparative Statics
p
s
=
a c bt
b +d
q
t
=
ad +bc bdt
b +d
p
b
= p
s
+t =
a c +dt
b +d
As t increases:
p
s
falls,
p
b
rises,
q
t
falls.
Competitive Markets
2.25
Comparative Statics
p
s
=
a c bt
b +d
q
t
=
ad +bc bdt
b +d
p
b
= p
s
+t =
a c +dt
b +d
As t increases:
p
s
falls,
p
b
rises,
q
t
falls.
Competitive Markets
2.26
Algebraic Incidence
p
s
=
a c bt
b +d
q
t
=
ad +bc bdt
b +d
p
b
=
a c +dt
b +d
The tax paid per unit by the buyer is
p
b
p
=
a c +dt
b +d

a c
b +d
=
dt
b +d
The tax paid per unit by the seller is:
p
p
s
=
a c
b +d

a c bt
b +d
=
bt
b +d
Competitive Markets
2.26
Algebraic Incidence
p
s
=
a c bt
b +d
q
t
=
ad +bc bdt
b +d
p
b
=
a c +dt
b +d
The tax paid per unit by the buyer is
p
b
p
=
a c +dt
b +d

a c
b +d
=
dt
b +d
The tax paid per unit by the seller is:
p
p
s
=
a c
b +d

a c bt
b +d
=
bt
b +d
Competitive Markets
2.27
Price Responses are Inversely Proportional to Elasticities
Demand response:
D

q
q
p
b
p
p
b
p

q p
D
q
Supply response:
S
=
q
q
p
s
p
p
s
p

q p
S
q
Competitive Markets
2.28
Tax Incidence and Relative Responsiveness
Dene tax incidence as:
I =
p
b
p
p
s
where:
p
b
p

q p
D
q
p
s
p

q p
S
q
Hence
I =
p
b
p
p
s

D
Competitive Markets
2.28
Tax Incidence and Relative Responsiveness
Dene tax incidence as:
I =
p
b
p
p
s
where:
p
b
p

q p
D
q
p
s
p

q p
S
q
Hence
I =
p
b
p
p
s

D
Competitive Markets
2.29
Tax Incidence with Perfect Elastic or Perfectly Inelastic Supply
Competitive Markets
2.30
Geometry of Tax Incidence
Competitive Markets
2.31
Tax Incidence and Responsiveness of Supply and Demand
The fraction of a $t quantity tax paid by buyers rises as supply
becomes more own-price elastic or as demand becomes less
own-price elastic.
When
D
= 0 and
S
> 0, buyers pay the entire tax, even though
it is levied on the sellers.
When
S
= 0 and
D
> 0, sellers pay the entire tax, even though
it is levied on the buyers.
Competitive Markets
2.32
Deadweight Loss
A quantity tax imposed on a competitive market reduces the quantity
traded and so reduces gains-to-trade (i.e. the sum of Consumers
and Producers Surpluses).
The lost total surplus is the taxs deadweight loss, or excess burden.
Competitive Markets
2.33
Tax Revenue
p
s
=
a c bt
b +d
q
t
=
ad +bc bdt
b +d
p
b
=
a c +dt
b +d
The total tax is then
T = tq
t
= t
ad +bc bdt
b +d
Note that this is a concave quadratic form. When bd > 0 there exists
a tax rate, t
which maximizes T. For t > t
, tax revenue decreases

with the tax rate.
Competitive Markets
2.33
Tax Revenue
p
s
=
a c bt
b +d
q
t
=
ad +bc bdt
b +d
p
b
=
a c +dt
b +d
T = tq
t
= t
ad +bc bdt
b +d
a tax rate, t

with the tax rate.
Competitive Markets
2.33
Tax Revenue
p
s
=
a c bt
b +d
q
t
=
ad +bc bdt
b +d
p
b
=
a c +dt
b +d
T = tq
t
= t
ad +bc bdt
b +d
a tax rate, t

with the tax rate.
Competitive Markets
2.34
Tax Revenue
Competitive Markets
2.35
Consumer and Producer Surplus
Competitive Markets
2.36
A Tax Affects Both Consumer and Producer Surplus
Competitive Markets
2.37
Deadweight Loss Measures Value of Trades which Disappear
Competitive Markets
2.38
Deadweight Loss is Zero When Quantities are Fixed
Competitive Markets
2.39
Deadweight Loss and Own-Price Elasticities
Deadweight loss due to a quantity tax rises as either market
demand or market supply becomes more own-price elastic.
If either
D
= 0 or
S
= 0 then the deadweight loss is zero.
Analysis of an economic policy proposal involves assessment of
both equity and efciency. In the Marshallian model, equity
impacts are evaluated on the basis of either (i) surplus
(consumer and producer) or (ii) tax incidence. Efciency in the
Marshallian model is assessed on the basis of the deadweight
loss.
Market Forces: Demand
and Supply
2a.1
Concepts
Lecture 2a
Market Forces: Demand and Supply
Marshallian Economics
September 30, 2011
ETH Zrich
and Supply
2a.2
Concepts
Overview
I. Market Demand Curve
- The Demand Function
- Determinants of Demand
- Consumer Surplus
II. Market Supply Curve
- The Supply Function
- Supply Shifters
- Producer Surplus
III. Market Equilibrium
IV. Price Restrictions
V. Comparative Statics
and Supply
2a.3
Concepts
Market Demand Curve
Shows the amount of a good that will be purchased at alternative
prices, holding other factors constant.
Law of Demand
The demand curve is downward sloping.
and Supply
2a.4
Concepts
Determinants of Demand
Income
Normal good
Inferior good
Prices of Related Goods
Prices of substitutes
Prices of complements
Advertising and consumer tastes
Population
Consumer expectations
and Supply
2a.5
Concepts
The Demand Function
A general equation representing the demand curve
Q
d
x
= f (P
x
, P
y
, M, H)
- Q
d
x
= quantity demand of good X.
- P
x
= price of good X.
- P
y
= price of a related good Y.
o Substitute good.
o Complement good.
- M = income.
o Normal good.
o Inferior good.
- H = any other variable affecting demand.
and Supply
2a.6
Concepts
Inverse Demand Function: Willingness to pay
- Price as a function of quantity demanded.
- Example:
- Demand Function
Q
d
x
(p) = 10 2p
- Inverse Demand Function:
P
x
(q) = 5 q/2
and Supply
2a.7
Concepts
Change in Quantity Demanded
and Supply
2a.8
Concepts
Change in Demand
and Supply
2a.9
Concepts
Consumer Surplus
The value consumers get from a good but do not have to pay for.
Consumer surplus will prove particularly useful in marketing and
other disciplines emphasizing strategies like value pricing and
price discrimination.
and Supply
2a.10
Concepts
I got a great deal!
That company offers a lot of bang for the
buck!
Amazon provides good value.
Total value greatly exceeds total amount
paid.
Consumer surplus is large.
and Supply
2a.11
Concepts
I got a lousy deal!
That car dealer drives a hard bargain!
I almost decided not to buy it!
They tried to squeeze the very last cent from
me!
Total amount paid is close to total value.
Consumer surplus is low.
and Supply
2a.12
Concepts
Consumer Surplus: Discrete Case
and Supply
2a.13
Concepts
Consumer Surplus: Continuous Case
and Supply
2a.14
Concepts
Market Supply Curve
The supply curve shows the amount of a good that will be
produced at alternative prices.
Law of Supply
- The supply curve is upward sloping.
and Supply
2a.15
Concepts
Supply Shifters
Input prices
Technology or government regulations
Number of rms
- Entry
- Exit
Substitutes in production
Taxes
- Excise tax
- Ad valorem tax
Producer expectations
and Supply
2a.16
Concepts
The Supply Function
An equation representing the supply curve:
Q
s
x
= f (P
x
, P
r
, W, H)
- Q
s
x
= quantity supplied of good X.
- P
x
= price of good X.
- P
r
= price of a production substitute.
- W = price of inputs (e.g., wages).
- H = other factors affecting supply.
and Supply
2a.17
Concepts
Inverse Supply Function: Willingness to accept
Price as a function of quantity supplied.
Example:
- Supply Function
Q
s
x
= 10 + 2P
x
- Inverse Supply Function:
P
x
(q) = 5 + 0.5q
and Supply
2a.18
Concepts
Change in Supply Quantity
and Supply
2a.19
Concepts
Change in Market Supply
and Supply
2a.20
Concepts
Producer Surplus
The amount producers receive in excess of the amount
necessary to induce them to produce the good.
and Supply
2a.21
Concepts
Market Equilibrium
The Price (P) that Balances supply and
demand
- Q
S
x
= Q
d
x
- No shortage or surplus
Steady-state
and Supply
2a.22
Concepts
If price is too low ...
and Supply
2a.23
Concepts
If price is too high ...
and Supply
2a.24
Concepts
Price Restrictions
Price Ceilings
- The maximum legal price that can be charged.
- Examples:
o Gasoline prices in the 1970s.
o Housing in New York City.
o Proposed restrictions on ATM fees.
Price Floors
- The minimum legal price that can be charged.
- Examples:
o Minimum wage.
o Agricultural price supports.
and Supply
2a.25
Concepts
Impact of a Price Ceiling
and Supply
2a.26
Concepts
Full Economic Price
The dollar amount paid to a rm under a price ceiling, plus the
non-pecuniary price.
P
F
= P
c
+
- P
F
= full economic price
- P
c
= price ceiling
- = nonpecuniary price
and Supply
2a.27
Concepts
An Example from the 1970s
Ceiling price of gasoline: $1.
3 hours in line to buy 15 gallons of gasoline:
- Opportunity cost: $5/hr.
- Total value of time spent in line: 3 ? $5 = $15.
- Non-pecuniary price per gallon: $15/15 = $1.
Full economic price of a gallon of gasoline: $1+$1=2.
and Supply
2a.28
Concepts
Impact of a Price Floor
and Supply
2a.29
Concepts
Comparative Static Analysis
How do the equilibrium price and quantity change when a
determinant of supply and/or demand change?
and Supply
2a.30
Concepts
Applications: Demand and Supply Analysis
Event: The WSJ reports that the prices of PC components are
expected to fall by 5-8 percent over the next six months.
Scenario 1: You manage a small rm that manufactures PCs.
Scenario 2: You manage a small software company.
and Supply
2a.31
Concepts
Use Comparative Static Analysis to see the Big Picture!
Comparative static analysis shows how the equilibrium price and
quantity will change when a determinant of supply or demand
changes.
and Supply
2a.32
Concepts
Scenario 1: Implications for a Small PC Maker
Step 1: Look for the Big Picture.
Step 2: Organize an action plan (worry about details).
and Supply
2a.33
Concepts
Big Picture: Impact of decline in component prices on PC market
and Supply
2a.34
Concepts
Big Picture Analysis: PC Market
Equilibrium price of PCs will fall, and equilibrium quantity of
computers sold will increase.
Use this to organize an action plan:
- contracts/suppliers?
- inventories?
- human resources?
- marketing?
- do I need quantitative estimates?
and Supply
2a.35
Concepts
Scenario 2: Software Maker
More complicated chain of reasoning to arrive at the Big Picture.
Step 1: Use analysis like that in Scenario 1 to deduce that lower
component prices will lead to
- a lower equilibrium price for computers.
- a greater number of computers sold.
Step 2: How will these changes affect the Big Picture in the
software market?
and Supply
2a.36
Concepts
Big Picture: Impact of lower PC prices on the software market
and Supply
2a.37
Concepts
Big Picture Analysis: Software Market
Software prices are likely to rise, and more software will be sold.
Use this to organize an action plan.
and Supply
2a.38
Concepts
Conclusion
Use supply and demand analysis to
- clarify the big picture (the general impact of a current event on
equilibrium prices and quantities).
- organize an action plan (needed changes in production,
inventories, raw materials, human resources, marketing plans,
etc.).
A Market Model with
Excel
Lecture 2b.1
Lecture Lecture 2b
A Market Model with Excel
September 30, 2011
ETH Zrich
A Market Model with
Excel
Lecture 2b.2
Calibrated Demand: Elasticity
The elasticity of demand (
D
> 0) is formally dene as:
D
=
% change quantity
% change price
=
Q
Q
D
P
D
P
D
This elasticity is a local approximation of the responsiveness of
quantity to price. The elasticity characaterizes the slope of the
demand function at a given price level.
A Market Model with
Excel
Lecture 2b.3
Linear Demand
We can use
D
together with a reference price and reference quantity
calibrate a demand function. The linear demand model based on
these data can be written as:
Q
D
=

Q
D
1 |
D
|
P
D
P
D
1
in which
Q
d
is the reference demand quantity
P
D
is the reference demand price
A Market Model with
Excel
Lecture 2b.4
Isoelastic Demand
The same input data can used to calibrate an isoelastic demand
function:
Q
D
=

Q
D
P
D
P
D
|
D
|
In the neighborhood of

P
D
, these functions are identical, yet as prices
depart from the reference point, the two functions may depart
signicantly.
A Market Model with
Excel
Lecture 2b.5
Digression: The Revenue Function
The revenue function at a given price is dened as:
R(p) = pQ(p)
Irregardless of the value of , revenue is a concave parabolic function
of price in the linear model. Let Q
denote the quantity for which

R(Q) is maximal. When < 1, maximal revenue occurs for Q
<

Q.
When > 1, Q
>

Q.
A Market Model with
Excel
Lecture 2b.6
Revenue Isoelastic Model
When
D
= 1, revenue is constant in the isoelastic model. Otherwise,
when < 1, Q
# 0, and when > 1, Q
" 1.
A Market Model with
Excel
Lecture 2b.7
Revenue Calculation Worksheet
A Market Model with
Excel
Lecture 2b.8
Inelastic Demand
A Market Model with
Excel
Lecture 2b.9
Elastic Demand
A Market Model with
Excel
Lecture 2b.10
A Coal Market Market
1 Find data on base year production, consumption and prices of
coal in a collection of countries which collectively represent
global coal supply and demand.
2 Calibrate a model to these data.
3 Perform counterfactural analysis by applying excise taxes in a
subset of regions, corresponding to the Annex-B member states.
4 Assume that coal supply is price elasticity (in the range of 1 to 2).
5 Assume that coal demand is price in-elastic (in the range of 0.5).
6 Evaluate the global leakage rate:
` =
% increase in coal use in non-Annex B states
% decrease in coal use in Annex B states
7 Does the leakage rate exceed 100% as is claimed by some
critical of climate policy?
8 Remember that The most interesting answer to any question in
economics is: It depends.
A Market Model with
Excel
Lecture 2b.10
` =
A Market Model with
Excel
Lecture 2b.10
` =
A Market Model with
Excel
Lecture 2b.10
` =
A Market Model with
Excel
Lecture 2b.10
` =
A Market Model with
Excel
Lecture 2b.10
` =
A Market Model with
Excel
Lecture 2b.10
` =
A Market Model with
Excel
Lecture 2b.10
` =
economics is:
It depends.
A Market Model with
Excel
Lecture 2b.10
` =
A Market Model with
Excel
Lecture 2b.11
Energy Data and Models
1 With the worldwide web, there are many data sources.
2 The data required for academic research is fundmentally
different than the data required by market participants, many of
whom are trying to shave small price differences.
3 Data is not very valuable without a model.
4 Economics offers several alternative approaches for modeling:
Econometrics works with large quantities of data and often very
few parametric assumptions.
Calibrated microeconomic models begin with an explicit theory and
relatively few data are required.
Econometrics can be concerned with measuring elasticities while
calibrated policy analysis seeks to assess the policy implications of
a given set of benchmark data values and elasticity assumptions.
5 Highschool students and naive undergraduates are typically
preoccupied with data. PhD students and profession
researchers are typically preoccpied with models.
A Market Model with
Excel
Lecture 2b.11
A Market Model with
Excel
Lecture 2b.11
A Market Model with
Excel
Lecture 2b.11
A Market Model with
Excel
Lecture 2b.11
A Market Model with
Excel
Lecture 2b.11
A Market Model with
Excel
Lecture 2b.11
A Market Model with
Excel
Lecture 2b.12
Supply Elasticity
The elasticity of supply (
S
> 0) is formally dened as:
S
=
% change quantity
% change price
=
Q
Q
s
P
s
P
s
The elasticity is a dimensionless representation of the slope of the
supply curve.
For calibrated policy analysis models, the elasticity of supply is a
model input. In many econometric exercies, the elasticity of supply is
a model output.
A Market Model with
Excel
Lecture 2b.12
Supply Elasticity
S
S
=
% change quantity
% change price
=
Q
Q
s
P
s
P
s
supply curve.
a model output.
A Market Model with
Excel
Lecture 2b.12
Supply Elasticity
S
S
=
% change quantity
% change price
=
Q
Q
s
P
s
P
s
supply curve.
a model output.
A Market Model with
Excel
Lecture 2b.13
Calibrated Linear Supply Functions
In calibrated equilibrium models we can use a reference price,
reference quantity and an elasticity of supply to dene a linear supply
function. That is, we can write:
Q
s
=

Q
s
1 +
S
P
s
P
s
1
where:
Q
s
is the reference supply quantity
P
s
is the reference supply price
S
is the price elasticity of supply
Note that when P
s
=

P
s
, Q
s
=

Q
s
.
A Market Model with
Excel
Lecture 2b.14
Calibrated Iso-Elastic Supply Functions
A simple alternative to the linear model is the iso-elastic model :
Q
s
=

Q
s
P
s
P
s
S
A Market Model with
Excel
Lecture 2b.15
A Simple Model of the Global Coal Market
The basic structure of the model is summarized by the equation:
X
r
S
r
(p) =
X
r
D
r
(p, t
r
)
in which
p is the world market price of coal
S
r
(p) is coal supply in region r .
t
r
is the specic tax on coal in region r .
D
r
(p, t
r
) is coal demand in region r .
The demand and supply functions employed in the model are linear,
hence:
S
r
(p) = a
r
+ b
r
p
and
D
r
(p, t
r
) =
r

r
(p + t
r
)
A Market Model with
Excel
Lecture 2b.16
Implementation in Excel
We will illustrate how this simple coal model can be implemented
in Excel.
The model consists of an Excel worksheet with regional data.
One cell in the sheet measures the equilibrium price.
Model benchmark inputs include base year supply, demand and
tax rates.
Model econometric inputs include elasticities of supply and
demand in each of the regions.
Model policy inputs include specic tax rates.
Model equilibrium is dened by a single variable: the
international coal price.
A model equilibrium determines supply and demand for each of
the regions.
A model equilibrium also determines the leakage rate.
A Market Model with
Excel
Lecture 2b.16
in Excel.
tax rates.
the regions.
A Market Model with
Excel
Lecture 2b.16
in Excel.
tax rates.
the regions.
A Market Model with
Excel
Lecture 2b.16
in Excel.
tax rates.
the regions.
A Market Model with
Excel
Lecture 2b.16
in Excel.
tax rates.
the regions.
A Market Model with
Excel
Lecture 2b.16
in Excel.
tax rates.
the regions.
A Market Model with
Excel
Lecture 2b.16
in Excel.
tax rates.
the regions.
A Market Model with
Excel
Lecture 2b.17
The Model Worksheet
The model worksheet is displayed below.
The market price variable is specied in B6 which has the
assigned range name P. This cell is used to dene equilibrium
demand and supply values in columns G and H.
The equilibrium values depend on the assigned policy
parameters, consumption tax rates which appear in column F.
The sum of squares market balance is dened as:
=
X
r
(S
r
D
r
)
2
This is displayed in cell B7.
If, for example, a tax rate is changed, then the model is out of
equilibrium and resulting imbalance is displayed in B7.
A Market Model with
Excel
Lecture 2b.17
The Model Worksheet
=
X
r
(S
r
D
r
)
2
A Market Model with
Excel
Lecture 2b.17
The Model Worksheet
=
X
r
(S
r
D
r
)
2
A Market Model with
Excel
Lecture 2b.17
The Model Worksheet
=
X
r
(S
r
D
r
)
2
A Market Model with
Excel
Lecture 2b.17
The Model Worksheet
=
X
r
(S
r
D
r
)
2
A Market Model with
Excel
Lecture 2b.18
The Model
A Market Model with
Excel
Lecture 2b.19
The Excel Solver
The model is solved using the Excel solver add-in (Tools >
Solver ...). In order to use the model, you may need to add a
reference to the Solver VBA add-in functions.
To use the solver we choose our target cell, the square market
imbalance, to be the "Target Cell" and choose the "Min" option
(see Figure 2 below).
Our only design variable is P, so the only cell we are going to
change is B6 (range name P).
Having specied these items, we click on the Solve button. The
model is solved instantaneously, and we are then presented with
a dialogue box asking whether to accept the solution (Figure 3).
A Market Model with
Excel
Lecture 2b.19
The Excel Solver
A Market Model with
Excel
Lecture 2b.19
The Excel Solver
A Market Model with
Excel
Lecture 2b.19
The Excel Solver
A Market Model with
Excel
Lecture 2b.20
The Solver Dialogue
A Market Model with
Excel
Lecture 2b.21
Accepting a Solution
A Market Model with
Excel
Lecture 2b.22
What is the insight?
Supply and demand elasticities for coal are low, and leakage rates
rarely exceed 10% for any sort of climate policies currently under
discussion.
A Market Model with
Excel
Lecture 2b.23
Quantitative Demand
Analysis
3a.1
Lecture 3a
Quantitative Demand Analysis
Elasticities and Estimation
October 7, 2011
ETH Zrich
Quantitative Demand
Analysis
3a.2
Overview
I. The Elasticity Concept
- Own Price Elasticity
- Elasticity and Total Revenue
- Cross-Price Elasticity
- Income Elasticity
II. Demand Functions
- Linear
- Log-Linear
III. Regression Analysis
Quantitative Demand
Analysis
3a.3
The Elasticity Concept
How responsive is variable G to a change in variable S
E
G,S
=
%G
%S
If E
G,S
> 0, then S and G are directly related.
If E
G,S
< 0, then S and G are inversely related.
If E
G,S
= 0, then S and G are unrelated.
Quantitative Demand
Analysis
3a.4
Formal Denition of Elasticity
An alternative way to measure the elasticity of a function
G = f (S) is
E
G,S
=
dG
dS
S
G
If E
G,S
> 0, then S and G are directly related.
If E
G,S
< 0, then S and G are inversely related.
If E
G,S
= 0, then S and G are unrelated.
Quantitative Demand
Analysis
3a.5
Own Price Elasticity of Demand
E
Q
x
,P
x
=
%Q
d
x
%P
x
Should be negative according to the law of demand.
Elastic:
|E
Q
x
,P
x
| > 1
Inelastic:
|E
Q
x
,P
x
| < 1
Unitary:
|E
Q
x
,P
x
| = 1
Quantitative Demand
Analysis
3a.6
Perfectly Elastic & Inelastic Demand
Quantitative Demand
Analysis
3a.7
Own-Price Elasticity and Total Revenue
Elastic
Increase (a decrease) in price leads to a decrease (an increase) in
total revenue.
Inelastic
Increase (a decrease) in price leads to an increase (a decrease) in
total revenue.
Unitary
Total revenue is maximized at the point where demand is unitary
elastic.
Quantitative Demand
Analysis
3a.8
Elasticity, Total Revenue and Linear Demand
Quantitative Demand
Analysis
3a.9
Quantitative Demand
Analysis
3a.10
Quantitative Demand
Analysis
3a.11
Quantitative Demand
Analysis
3a.12
Quantitative Demand
Analysis
3a.13
Quantitative Demand
Analysis
3a.14
Quantitative Demand
Analysis
3a.15
Quantitative Demand
Analysis
3a.16
Demand, Marginal Revenue (MR) and Elasticity
For a linear inverse demand function,
MR(Q) = a +2bQ,
where b < 0.
When
MR > 0, demand is elastic;
MR = 0, demand is unit elastic;
MR < 0, demand is inelastic.
Quantitative Demand
Analysis
3a.17
Factors Affecting the Own-Price Elasticity
Available Substitutes
The more substitutes available for the good, the more elastic the
demand.
Time
Demand tends to be more inelastic in the short term than in the
long term.
Time allows consumers to seek out available substitutes.
Expenditure Share
Goods that comprise a small share of consumers budgets tend to
be more inelastic than goods for which consumers spend a large
portion of their incomes.
Quantitative Demand
Analysis
3a.18
Cross-Price Elasticity of Demand
E
Q
x
,P
y
=
%Q
d
x
%P
y
If E
Q
x
,P
y
> 0 then X and Y are substitutes.
If E
Q
x
,P
y
< 0 then X and Y are complements.
Quantitative Demand
Analysis
3a.19
Predicting Revenue Changes from Two Products
Suppose that a rm sells two related goods, X and Y. If the price of
X is change, then total revenue will change by:
R =
R
X
(1 +E
Q
x
,P
x
) +R
Y
E
Q
y
,P
x
%P
X
Quantitative Demand
Analysis
3a.20
Income Elasticity
E
Q
x
,M
=
%Q
d
x
%M
If E
Q
x
,M
> 0, then X is a normal good.
If E
Q
x
,M
< 0, then X is a inferior good.
Quantitative Demand
Analysis
3a.21
Uses of Elasticities
Pricing.
Managing cash ows.
Impact of changes in competitors prices.
Impact of economic booms and recessions.
Impact of advertising campaigns.
And lots more!
Quantitative Demand
Analysis
3a.22
Example 1: Pricing and Cash Flows
According to an FTC Report by Michael Ward, AT&Ts own price
elasticity of demand for long distance services is -8.64.
AT&T needs to boost revenues in order to meet its marketing
goals.
To accomplish this goal, should AT&T raise or lower its price?
Quantitative Demand
Analysis
3a.23
Answer: Lower price!
Since demand is elastic, a reduction in price will increase
quantity demanded by a greater percentage than the price
decline, resulting in more revenues for AT&T.
Quantitative Demand
Analysis
3a.24
Example 2: Quantifying the Change
If AT&T lowered price by 3 percent, what would happen to the
volume of long distance telephone calls routed through AT&T?
Quantitative Demand
Analysis
3a.25
Answer: Calls Increase!
Calls would increase by 26 percent!
E
Q
x
,P
x
= 8.64 =
%Q
d
x
%P
x
8.64 =
%Q
d
x
3%
) %Q
d
x
= 26%
Quantitative Demand
Analysis
3a.26
Example 3: Impact of a Change in a Competitors Price
According to an FTC Report by Michael Ward, AT&Ts cross
price elasticity of demand for long distance services is 9.06.
If competitors reduced their prices by 4 percent, what would
happen to the demand for AT&T services?
Quantitative Demand
Analysis
3a.27
Answer: AT&Ts Demand Falls!
AT&Ts demand would fall by 36 percent!
E
Q
x
,P
y
= 9.06 =
%Q
d
x
%P
y
9.06 =
%Q
d
x
4%
) %Q
d
x
= 36%
Quantitative Demand
Analysis
3a.28
Interpreting Demand Functions
Mathematical representations of demand curves.
Example:
Q
d
x
= 10 2P
x
+3P
y
2M
Law of demand holds (coefcient of P
x
is negative).
X and Y are substitutes (coefcient of P
y
is positive).
X is an inferior good (coefcient of M is negative).
Quantitative Demand
Analysis
3a.29
Linear Demand Functions and Elasticities
General Linear Demand Function and Elasticities:
Q
d
x
=
0
+
x
P
x
+
y
P
y
+
M
M +
H
H
Own-Price Elasticity:
E
Q
x
,P
x
=
x
P
x
Q
x
Cross-Price Elasticity:
E
Q
x
,P
y
=
y
P
y
Q
x
Income Elasticity:
E
Q
x
,M
=
M
M
Q
x
Quantitative Demand
Analysis
3a.30
Example of Linear Demand
Q
d
= 10 2P
Own-Price Elasticity: (2)P/Q.
If P = 1, Q = 8 (since 10 2 = 8).
Own price elasticity at P = 1, Q = 8:
(2)(1)/8 = 0.25
.
Quantitative Demand
Analysis
3a.31
Log-Linear Demand
General Log-Linear Demand Function:
lnQ
d
x
=
0
+
x
P
x
+
y
P
y
+
M
M +
H
H
Own-Price Elasticity:
x
Cross-Price Elasticity:
y
Income Elasticity:
M
Quantitative Demand
Analysis
3a.32
Example of Log-Linear Demand
ln(Q
d
) = 10 2ln(P).
Own Price Elasticity: 2.
Quantitative Demand
Analysis
3a.33
Graphical Representation of Linear and Log-Linear Demand
Quantitative Demand
Analysis
3a.34
Regression Analysis
One use is for estimating demand functions.
Important terminology and concepts:
Least Squares Regression model: Y = a + bX + e.
Least Squares Regression line:

Y =

a +

bX
Condence Intervals.
t-statistic.
R-square or Coefcient of Determination.
F-statistic.
Quantitative Demand
Analysis
3a.35
An Example
We can use a spreadsheet to estimate the following log-linear
demand function.
Q
x
=
0
+
x
lnP
x
+e
Quantitative Demand
Analysis
3a.36
Summary Output
Quantitative Demand
Analysis
3a.37
Interpreting the Regression Output
The estimated log-linear demand function is:
ln(Q
x
) = 7.58 0.84ln(P
x
).
Own price elasticity: 0.84 (inelastic).
How good is our estimate?
t-statistics of 5.29 and -2.80 indicate that the estimated coefcients
are statistically different from zero.
R-square of 0.17 indicates the ln(P
x
) variable explains only 17
percent of the variation in ln(Q
x
).
F-statistic signicant at the 1 percent level.
Quantitative Demand
Analysis
3a.38
Conclusion
Elasticities are tools you can use to quantify the impact of
changes in prices, income, and advertising on sales and
revenues.
Given market or survey data, regression analysis can be used to
estimate:
Demand functions.
Elasticities.
A host of other things, including cost functions.
Managers can quantify the impact of changes in prices, income,
advertising, etc.
Demand Estimation in
Economics
3b.1
Lecture 3b
Demand Estimation in Economics
Intro to Econometrics
October 7, 2011
ETH Zrich
Economics
3b.2
Econometric modeling
Economists use two main type of statistical models to forecast
and provide policy analysis.
1 Single-equation models study a variable of interest with a single
(linear or non-linear) function of a number of explanatory variables.
2 In multiple or simultaneous equation models, the variable of
interest is a function of several explanatory variables which are
related to each other with a set of equations.
Specic estimation techniques may be needed depending on the
data type:
1 A times series is a time-ordered (daily, weekly, . . . ) sequence of
data (price, income, . . . ) which often requires special statistical
treatment.
2 a cross section refers to data collected by observing many
subjects (individuals, rms or countries) at the same point in time.
Its analysis usually consists of comparing the differences among
the subjects.
Here we provide some background on demand estimation and
regression analysis in the context of a single-equation approach.
Economics
3b.3
Simple Linear Demand Estimation
"Nobody employs expensive, time-consuming and complicated
demand estimation techniques when inexpensive and simple
methods work just ne.", Hirschey (2009, p.162).
Example 1: Grasshopper (GZ), one of Zurichs soccer teams
playing in the Swiss Soccer Super League, offered CHF 5 off the
CHF 20 regular price of reserved seats. Sales increased from
6000 to 7000 seats per game. What is the demand for GZs
game tickets? Assuming a linear relationship:
Q = a + bP )
(
6000 = a + b(20)
7000 = a + b(15)
Solving for a and b gives the deterministic demand relationship:
Q = 10000 + -200P
| {z }
demand
or, equivanlently, P = 50 +0.005Q
| {z }
inverse demand
(1)
Economics
3b.4
Price Elasticity of Demand
From Example 1, we notice that the slope of the demand function
being negative, GZs games are a normal good!
We can also compute the price elasticity resulting directly from
the price change (arc elasticty):
Q
1
Q
0
Q
0
P
1
P
0
P
0
=
70006000
6000
1520
20
=
2
3
Note that in the context of a linear function, the arc elasticty is
equal to the point elasticity:
@Q
@P
P
0
Q
0
= 200
20
6000
=
2
3
Economists usually plot the inverse demand, i.e., the price
variable is on the y-axis. The inverse demand function is useful
in several contexts.
Inverse Linear Demand Function
Economists usually plot demand functions with the price variable
is on the y-axis and the quantities in the x-axis:
Economics
3b.6
Revenue-maximizing output level
If the cost of producing an additional soccer game for GZ is xed,
we can use the inverse ticket demand function (1) to nd the
revenue-maximizing price level :
T = P Q = (50 0.005Q
| {z }
inverse demand
) = 50Q 0.005Q
2
Lets maximize T with respect to Q:
FOC: (@T/@Q) = 0 )50 0.01Q = 0 )Q
= 5000
SOC: (@
2
T/@Q
2
) < 0 )0.01 < 0 )Q
is max!
Price at Q
: P
= 50 0.005(5000) = 25
Verify on slide 3 that at P = 20, T = 120000. Reducing the price
to P = 15 (-25%) increased the ticket sales in a lower proportion
(+16.6%) to 6000. Therefore, T dropped to 105000. Setting the
price to 25 could have generated 125000 in ticket revenues.
Would " P have been judicious for GZ? Well, less costumers
(-1000) means less high margin products (sodas, beers,
burgers,. . . ) sold!
Identication Problem
Estimating demand relations can be complicated because of the interplay
between demand and supply.
The dashed AB line is not a demand. Advanced statistical techniques are
required to identify demand in that case.
C. Ordas Criado (CEPE-ETH) Managerial Economics (Fall 2010) Demand Estimation 9 / 40
Statistical Relation
A deterministic relation is an association between variables that is known
with certainty.
Economic relationships are not deterministic in nature because they cannot
be predicted with absolute accuracy.
Real world economic data are rather of statistical type :
Linear Models
A statistical model in the context of demand estimation for good x coud be
of the form :
Q
x
= a
0
+ a
1
P
x
+ a
2
m + a
3
P
y
+ (3)
Q
x
= b
0
P
b
1
x
m
b
2
P
b
3
y
e
(4)
where and are random terms that follow some statistical distribution.
Equation (3) is clearly linear. Some nonlinear functions, such as (4), are
linear in the parameters. To see why, note that:
log Q
x

Q
x
= log b
0

0
+b
1
log P
x

P
x
+b
2
log m

m
+b
3
log P
y

P
y
+
Q
x
=
0
+ b
1
P
x
+ b
2
m + b
3
P
y
+ (5)
The parameters of model (4) could be estimated with the linear model (5).
The most popular technique to estimate the coecients of functional forms
which are linear in the parameters is linear regression.
Linear Regression
Linear regression consists in nding the best-tting line that minimizes the
sum of squared deviations between the regression line and the set of
original data points. This technique is also know as the Ordinary Least
Squares (OLS) method.
Ordinary Least Squares (OLS)
Consider the following multiple regression model:
y
i
=
0
+
1
x
i 1
+ . . . +
p
x
ip
+ (6)
with n observations (i = 1, 2, . . . , n), p explanatory variables and K = p + 1
coecents (the
p
s plus the intercept
0
, where k = 0, 1, 2, . . . , K).
The OLS method nds the parameters (called

) such that :
min
0
,
1
,...,
p
n
i =1
(
i
)
2
=
n
i =1
(y
i

0
1
x
i 1
. . .
1
x
ip
)
2
(7)
Problem (7) has a closed form and unique solution when the explanatory
variables are linearly independent, i.e., no exact linear relationships exist
between two or more explanatory variables.
Most statistical softwares possess pre-implemented routines/functions to
perform regression analysis (Excel, Matlab, R, SPSS, S-Plus, Stata, . . . )
Fundamental OLS Assumptions
Four fundamental assumptions are necessary to get unbiased estimates of
the parameters and to carry statistical inference with a regression model:
1
the model is correctly specied, i.e., the relationship is linear in the
regression parameters .
2
each term
i
comes from a normal distribution with mean 0 and
constant variance
2
and it is independent of each other;
3
the explanatory variables x
1
, x
2
, . . . , x
p
are nonrandom, measured
without errors and independent of each other and of the intercept;
4
the error
i
is uncorrelated with the observations x
ip
for all p.
These assumptions can be formally veried (out-of-scope of this lecture). If
they are plausible, you can interpret the regression results.
Correlation Coecient
The goodness of t of the regression estimates must be evaluated before
interpreting the regression coecients.
The most straightforward measure is simply the correlation coecient
between the y data and their tted counterpart, called y:
R = cor (y, y) =
(y
i
y)(y
i

y)
2
(y
i
y)
2

( y
i

y
i
)
2
(8)
where y is the mean of the y
i
s and

y is the mean of the tted values ( y
i
s).
Note that R [0, 1]. The closer R is to 1, the better the t.
Explained and Unexplained Variation of Regression Fits
Other important goodness of t measures (the R
2
and the F-statistic) rely
on a decomposition of the variation of the dependent variable y into total,
explained and unexplained variation:
SST =
n
i =1
(y
i
y)
2
sum of squared deviations in y total variation (9)
SSR =
n
i =1
( y
i
y)
2
sum of squares of regression explained variation (10)
SSE =
n
i =1
(y
i
y
i
)
2
sum of squared errors unexplained variation (11)
where y
i
is the regression estimate of y
i
and y is the mean of the y
i
s.
It is not dicult to show that SST = SSR + SSE.
Explained and Unexplained Variation of Regression Fits
Goodness-of-Fit of the Regression Line: R
2
The R
2
captures the proportion of total variation of the dependent variable
y explained by the full set of independent variables and it is dened as
R
2
=
SSR
SST
= 1
SSE
SST
. (12)
The R
2
in (12) is equal to the square of R in (8) only when regression (6)
includes an intercept. The closer the R
2
is to 1, the larger the share of
variation explained by the model.
Note that adding explanatory variables to the regression never penalizes the
R
2
.
The R
2
can be compared across models as long as the y variable shares the
same units of measurement.
Goodness-of-Fit of the Regression Line: Adjusted R
2
A downward-adjusted version of the R
2
, called adjusted R
2
, exists to
account for the degrees of freedom, i.e. the number of observations
beyond the minimum needed to calculate the regression statistic. The
adjusted R
2
is
R
2
adj
= 1
SSE/(n K)
SST/(n 1)
= 1
n 1
n K
(1 R
2
). (13)
Note that R
2
adj
is not the share of total variance explained by the regression
model (it can be negative even in the presence of an intercept).
Preference should be given to the R
2
adj
when comparing regression models
with dierent number of predictors.
The closer R
2
adj
to 1, the better the model.
Global Signicance of the Regressors - the F-test
The F-statistic tells if the explanatory variables as a group explain a
statistically signicant share of the variation in the dependent variable :
F =
SSR/K 1
SSE/(n K)
=
MSR
MSE
=
R
2
/(K 1)
(1 R
2
/(n K))
(14)
MSR = (SSR/K 1) is also called Mean Squares of Regression and
MSE = (SSE/n K) is the Mean Squared Errors. The term df 1 = K 1
corresponds to the numerators degrees of freedom while df 2 = n K is the
denominators degrees of freedom.
Note that F 0. If R
2
= 0, then F = 0 and y is statistically unrelated to x
variables.
Data series always display some (weak) statistical relationships.
How large should F be to ensure that at least some of the explanatory
variables explain a statistically signicant portion of the variation in y ?
Building the F-distribution
F is a random variable whose statistical distribution can be determined
under some assumptions.
Recall from equation (14) that F depends on the tted values of the
regression model (through the SS terms) and on two dierent numbers of
degrees of freedom.
Under the assumptions that :
1
the regression errors are normally distributed (see slide 14),
2

1
=
2
= . . . =
p
= 0 in regression (6),
we can get statistical distributions of F, called F-distributions, which
depend on the two numbers of degrees of freedom.
Assumption (2) above is the null hypothesis under which the F-distribution
is derived. It assumes that none of the explanatory variables x has a
signicant relationship with y.
The F-distributions are in general highly skewed to the right and they
become more symmetric as the sample size increases.
F-distributions
The F-distribution depends on the degrees of freedom.
0 2 4 6 8 10
0
.
0
0
.
2
0
.
4
0
.
6
0
.
8
1
.
0
1
.
2
F
D
e
n
s
i
t
y
F(df1=1,df2=10,alpha=5%)
F-statistics above the colored thresholds suggest signicant contribution of
the explanatory variables at the 5% signicance level.
Thresholds for the F-test
The size of the F-statistic from equation (14) is then compared to the
F-values derived under the null hypothesis. If the F-statistic lies far in the
right tail of the F-distribution, the null hypothesis is unlikely to be true for
the investigated dataset.
Statisticians usually consider that a F-statistic which has only 5% chances
(or lower) to be observed under the null hypothesis is sucient evidence to
reject the null hypothesis. This rejection level is called signicance level
and it is noted .
Statistical tables of F-distributions exist for dierent levels. They
provide critical F-values, noted F
(df 1,df 2,)

, for a large range of degrees of
freedom. They report P(F > F
(df 1,df 2)
) = . To reject the null hypothesis
at the signicance level of = 5%, the following criteria must hold:
F > F
(df 1,df 2,0.05)

P(F) < 0.05 (15)
The term P(F) in (15), called the p-value of the F computed with (14),
corresponds to the probability that a F-statistic at least as extreme as F is
observed under the null hypothesis. Both criteria in (15) are identical.
Precision of the Regression Coecients
Rejecting the null hypothesis of the F-test ensures that the regressions
predictors as a whole contribute to explain a statistically signicant portion
of the variation in the dependent variable y. We can then proceed to
analyze the relationship between each explanatory variable and y.
Before interpreting their sign and magnitude, the precision and
reliability of each individual coecient can be assessed with the help of:
1
its standard error or standard deviation, denoted se
k
;
2
its t-statistic t =

k
/se
k
.
The detailed calculation of se
k
is not shown here (it is part of any standard
regression output).
When the size of a coecient (or some deviation from it) is large as
compared to its standard deviation, the relationship between x
k
and y is
expected to be strong.
Testing the Regression Coecients: 2-tail or 1-tail t-tests
Two-tailed t-test: If we want to assess whether an individual coecient
k
is signicantly dierent from some arbitrary (possibly null)
, we can
derive the theoretical distribution of t =

se
k
under the null hypothesis
that

k
=
. We then construct an interval around
(called condence
interval) which contains with probability 1 the true value
. If the
t-statistic that we obtain from the regression with the observed data does
not fall within the condence interval, we reject the null in favor of the
alternative

k
=
.
One-tailed t-test: Other alternative hypotheses can be of interest, in
particular,

k
>
or

k
<
at a signicance level of . Such tests simply

require the absolute value of t =

se
k
to be larger than some theoretical
threshold.
The appropriate distribution for the one-tailed or two-tailed t-tests when the
OLS assumptions (slide 14) hold is the Students t-distribution. The
related critical value is denoted t
(nK,)
and depends on the degrees of
freedom n K.
Student distribution
The t-statistic can be shown to be distributed as a Students t-distribution
centered on
(here below
= 0):
10 5 0 5 10
0
.
0
0
.
1
0
.
2
0
.
3
0
.
4
t
D
e
n
s
i
t
y
stud(df=1,alpha=5%)
stud(df=3,alpha=5%)
norm(0,1,alpha=5%)
The colored dots are two-tails critical values for = 5%. Note that the
t-distribution tends toward the Normal shape as n K increases.
Signicance Level of the Regression Coecients
Again, we can rely indierently on either critical values of the t-statistic,
noted t
(nK,)
, or on a p-value of the t-statistic.
For testing the null hypothesis that
k
=
at the signicance level

against the alternative
k
=
(2-tail t-test):
compute t =

se
k
;
if |t| > t
(nK,/2)
or if P(|t|) < , reject the null hypothesis in favor of
the alternative of signicant dierence at the signicance level of .
For testing the null hypothesis that
k
=
against the alternative

k
>
or
k
<
at the signicance level (1-tail t-test),

use the former t ratio
if |t| > t
(nK,)
or P(|t|) < , reject the null hypothesis in favor of the
chosen unilateral alternative at the signicance level .
Tables of the t-distribution may report 1-tail p-values 1 P(t t
nK,
) =
or 2-tail p-values 1 P(|t
nK,/2
| t) = . Be aware of what you use.
Condence Interval around the Regression Coecients
Condence intervals at the 1 level can also be constructed around

k
.
If you use a table of the t-distribution (2-tail t-test):
k
t
(nK,/2)
se
k
(16)
If that interval does not include some arbitrary (and possibly null) value
,
the regression coecient is signicantly dierent from
at the
signicance level.
Once you have carried out the appropriate individual t-tests on the

k
s, you
can proceed to interpret the coecients.
Interpreting Regression Coecients
Regression coecients are parameters of a functional relationship, so they
are straightforward to interpret!
For the linear demand function :
Q =

0
+

P +

2
m + . . .

1
=

Q
P
;

2
=

Q
m
. . . (17)
1
is the change in

Q corresponding to a unit change in

P when all other
explanatory variables are kept constant.
If the variables are in logarithms, e.g.

Q = log Q,

P = log P, m = log m in
equation (17), remember that the coecients are elasticities:
1
=
log Q
log P
=
1
Q
Q
1
P
P
=
Q
Q
P
P
; . . . (18)
Note that when you have more than one explanatory variable in a regression,
the regression coecients are partial regression coecients, i.e.,
1
= cor(
Q,

P) in equation (17).
A Regression Example with Excel 2007/2010
To replicate this example, use the le regression.xlsx from the course
website. These data are from Hirschey (2009, P.190).
We estimate the following single equation demand model:
UNIT SOLD =
0
+
1
PRICE +
2
ADVERT +
3
PERS SELL + (19)
For performing regression analysis with Excel 2007/2010, you need rst to
enable Excels Data Analysis Toolbox:
1
go the the File tab or click the Oce button and then click on Options
2
click on Add-Ins, select Analysis Toolpak in the Inactive Application
Add-ins and click on the Go. . . button
3
The Add-Ins window will pop up. Select Analysis Toolpak and click
OK
You can check that the Data Analysis Toolbox has been properly enabled by
selecting the Data tab in Excel and checking that the Data Analysis option
is available under the Analysis buttons.
Then open regression.xlsx in Excel and use the data in the data sheet.
A Regression Example with Excel - Steps (1) and (2)
Select Analysis Toolpak and click on the Go. . . button
A Regression Example with Excel - Step (3)
Select Analysis Toolpak and press OK
A Regression Example with Excel - Regression Tool
Click on the Data Analysis button and select Regression
A Regression Example with Excel - Regression Window
To replicate the results, use the same options in the Regression Window.
A Regression Example with Excel - Regression Output
A Regression Example with Excel - Regression Output
The Excel regression output generated above is divided in 4 main parts:
1
Regression statistics (R, R
2
, R
2
adj
, se
reg
, obs.)
2
SST, SSR, SSE and F-test, called (Analysis of Variance or ANOVA)
3
Regression Coecients
4
Residuals
The link between the Excel output and the formulas from the former slides is
emphasized below. The sheet regression (2) in regression.xlsx provides
further formulas checks (in yellow) that can be of interest.
A Regression Example with Excel - Regression Statistics
Lets focus on the Regression statistics:
The multiple correlation coecient R = cor (y, y) = 0.98 is very high.
This is not too surprising when time-series are employed.
The R square indicates that the regression explains 97% of the total
variance. It can be computed either by squaring the above R (because
the regression includes a constant: 0.98
2
= 0.97) or with information
from the ANOVA table (try to apply equation (12)).
In regressions based on cross-sectional data, R
2
> 0.5 is already a good
tting performance.
The R
2
adj
= 0.958 is pretty close to the R
2
which indicates that the
penalization linked to the degrees of freedom is not large.
The regression standard error (123.92) corresponds to the denominator
in equation (14). You can check this by typing in an Excel cell:
=sqrt(sumsq(resid range)/(12-4))
(replacing resid range with the appropriate range, check the sheet
regression (2)).
A Regression Example with Excel - ANOVA Table
The ANOVA table :
The gures reported in column Sum of Squares (SS) correspond to
SSR, SSE and SST from equations (9) to (11).
The MSR (Mean Square Regression) and MSE (Mean Square Errors)
gures correspond to numerator and denominator from equation (14)
and F = MSR/MSE = 85.40.
To judge if F is large enough (is the contribution of all the predictors
to the explained variation is signicant?), we can check whether or not
P(F) < 0.05: Signicance F being very small, we reject the null
hypothesis at the 5% signicance level and conclude that the predictors
(price, advertising and personal sells) contribute to explain the
variation of quantity sold.
Note that you can get the P(F) with the following Excel function:
=fdist(F,df1,df2)
(replace the F, df1, df2 with appropriate information from the
ANOVA table, check the sheet regression (2))
A Regression Example with Excel - Coecients Results
Regarding the regression coecients:
The p-value of the t-stat is lower than 5% for most coecients, which
means that they are signicantly dierent from 0 at that signicance
level. You can check that the T-stat column is the Coecients
column divided by the Standard error column.
The Excel functions that provides the 1-tail or 2-tail p-values of the
Students t-distribution is tdist() and the one for getting t
(nK,)
is
t.inv().
We can test if the price coecient (-0.296) is signicantly lower than
0 by simply comparing |t| = 2.908 with the unilateral cuto
t
124,0.05
= 1.86 as indicated in slide 27. As |t| > t
we reject the null

in favor of a signicantly negative coecient.
We also notice that for CHF 100 spent in Advertising we get an
average of 3.6 units sold/month.
Finally note that Excel provides 95% condence intervals around the
coecients. They correspond to those described in equation (16).
References
Hirschey M., Managerial Economics, 12th Edition, Ch.5.
Chatterjee S., Hadi A., Price B., Regression Analysis by Example, 3rd
Edition, Ch.3.
W. Greene, Econometric Analysis, 6th Edition, Ch.3.
Theory of Individual Behavior
October 14, 2011
Overview
I. Consumer Behavior
Indifference Curve Analysis.
Consumer Preference Ordering.
II. Constraints
The Budget Constraint.
Changes in Income.
Changes in Prices.
III. Consumer Equilibrium
IV. Indifference Curve Analysis & Demand Curves
Individual Demand.
Market Demand.
Consumer Behavior
Consumer Opportunities
The possible goods and services consumer can
afford to consume.
Consumer Preferences
The goods and services consumers actually
consume.
Given the choice between 2 bundles of goods
a consumer either:
Prefers bundle A to bundle B: A B.
Prefers bundle B to bundle A: A B.
Is indifferent between the two: A B.
Indifference Curve Analysis
Indifference Curve
A curve that defines the
combinations of 2 or more goods
that give a consumer the same
level of satisfaction.
Marginal Rate of
Substitution
The rate at which a consumer is
willing to substitute one good for
another and maintain the same
satisfaction level.
I.
II.
III.
Good Y
Good X
Consumer Preference Ordering
Properties
Completeness
More is Better
Diminishing Marginal Rate of Substitution
Transitivity
Complete Preferences
Completeness Property
Consumer is capable of expressing
preferences (or indifference)
between all possible bundles. (I
dont know is NOT an option!)
If the only bundles available to
a consumer are A, B, and C,
then the consumer
is indifferent between A and
C (they are on the same
indifference curve).
will prefer B to A.
will prefer B to C.
I.
II.
III.
Good Y
Good X
A
C
B
More Is Better!
More Is Better Property
Bundles that have at least as much of
every good and more of some good
are preferred to other bundles.
Bundle B is preferred to A since B
contains at least as much of
good Y and strictly more of good
X.
Bundle B is also preferred to C
since B contains at least as much
of good X and strictly more of
good Y.
More generally, all bundles on
IC
III
are preferred to bundles on
IC
II
or IC
I
. And all bundles on IC
II

are preferred to IC
I
.
I.
II.
III.
Good Y
Good X
A
C
B
1
33.33
100
3
Diminishing MRS
MRS
The amount of good Y the consumer is
willing to give up to maintain the same
satisfaction level decreases as more of
good X is acquired.
The rate at which a consumer is willing to
substitute one good for another and
maintain the same satisfaction level.
To go from consumption bundle A to
B the consumer must give up 50 units
of Y to get one additional unit of X.
To go from consumption bundle B to
C the consumer must give up 16.67
units of Y to get one additional unit
of X.
To go from consumption bundle C to
D the consumer must give up only
8.33 units of Y to get one additional
unit of X.
I.
II.
III.
Good Y
Good X
1 3 4 2
100
50
33.33
25
A
B
C
D
Consistent Bundle Orderings
Transitivity Property
For the three bundles A, B, and C,
the transitivity property implies
that if C B and B A, then C
A.
Transitive preferences along with
the more-is-better property imply
that
indifference curves will not
intersect.
the consumer will not get
caught in a perpetual cycle of
indecision.
I.
II.
III.
Good Y
Good X
2 1
100
5
50
7
75
A
B
C
The Budget Constraint
Opportunity Set
The set of consumption bundles that
are affordable.
P
x
X + P
y
Y M.
Budget Line
The bundles of goods that exhaust a
consumers income.
P
x
X + P
y
Y = M.
Market Rate of Substitution
The slope of the budget line
-P
x
/ P
y
.

Y
X
The Opportunity Set
Budget Line
Y = M/P
Y
(P
X
/P
Y
)X
M/P
Y
M/P
X
Changes in the Budget Line
Changes in Income
Increases lead to a parallel,
outward shift in the budget
line (M
1
> M
0
).
Decreases lead to a parallel,
downward shift (M
2
< M
0
).
Changes in Price
A decreases in the price of
good X rotates the budget
line counter-clockwise (P
X
0
>
P
X
1
).
An increases rotates the
budget line clockwise (not
shown).

X
Y
X
Y
New Budget Line for
a price decrease.
M
0
/P
Y
M
0
/P
X
M
2
/P
Y
M
2
/P
X
M
1
/P
Y
M
1
/P
X
M
0
/P
Y
M
0
/P
X
0
M
0
/P
X
1
Consumer Equilibrium
The equilibrium
consumption bundle
is the affordable
bundle that yields the
highest level of
satisfaction.
Consumer equilibrium
occurs at a point where
MRS = P
X
/ P
Y.
Equivalently, the slope of
the indifference curve
equals the budget line.
I.
II.
III.
X
Y
Consumer
Equilibrium
M/P
Y
M/P
X
Price Changes and Consumer
Equilibrium
Substitute Goods
An increase (decrease) in the price of good X leads to an
increase (decrease) in the consumption of good Y.
Examples:
Coke and Pepsi.
Verizon Wireless or AT&T.
Complementary Goods
An increase (decrease) in the price of good X leads to a
decrease (increase) in the consumption of good Y.
Examples:
DVD and DVD players.
Computer CPUs and monitors.
Complementary Goods
When the price of
good X falls and the
consumption of Y
rises, then X and Y
are complementary
goods. (P
X
1
> P
X
2
)
Pretzels (Y)
Beer (X)
II
I
0
Y
2
Y
1
X
1
X
2
A
B
M/P
X
1
M/P
X
2
M/P
Y
1
Income Changes and Consumer
Equilibrium
Normal Goods
Good X is a normal good if an increase
(decrease) in income leads to an increase
(decrease) in its consumption.
Inferior Goods
Good X is an inferior good if an increase
(decrease) in income leads to a decrease
(increase) in its consumption.
Normal Goods
An increase in
income increases
the consumption of
normal goods.
(M
0
< M
1
).
Y
II
I
0
A
B
X
M
0
/Y
M
0
/X
M
1
/Y
M
1
/X
X
0
Y
0
X
1
Y
1
Decomposing the Income and
Substitution Effects
Initially, bundle A is consumed. A
decrease in the price of good X
expands the consumers opportunity
set.
The substitution effect (SE) causes the
consumer to move from bundle A to B.
A higher real income allows the
consumer to achieve a higher
indifference curve.
The movement from bundle B to C
represents the income effect (IE). The
new equilibrium is achieved at point
C.
Y
II
I
0
A
X
C
B
SE
IE
A Classic Marketing Application
Other
goods
(Y)
II
I
0
A
C
B F
D
E
Pizza
(X)
0.5 1 2
A buy-one,
get-one free
pizza deal.
Individual Demand Curve
An individuals
demand curve is
derived from each
new equilibrium point
found on the
indifference curve as
the price of good X is
varied.
X
Y
$
X
D
II
I
P
0

P
1
X
0
X
1
Market Demand
The market demand curve is the horizontal
summation of individual demand curves.
It indicates the total quantity all consumers would
purchase at each price point.
Q
$ $
Q
50
40
D
2
D
1

Individual Demand
Curves
Market Demand Curve
1 2 1 2 3
D
M

Conclusion
Indifference curve properties reveal information
about consumers preferences between bundles of
goods.
Completeness.
More is better.
Diminishing marginal rate of substitution.
Transitivity.
Indifference curves along with price changes
determine individuals demand curves.
Market demand is the horizontal summation of
individuals demands.

Optimization Tools
4.1
Lecture 4
Optimization Tools
Lagrangian Methods
October 14, 2011
ETH Zrich
Optimization Tools
4.2
Good Mathematical References for Economics
Mathematics for Economists by Carl P. Simon and Lawrence
Bloom, Norton, 1994. (an essential reference)
Optimization in Economic Theory by Avinash K. Dixit, Oxford,
1975. (a sentimental favorite)
Mathematical methods for economic theory: a tutorial by Martin
J. Osborne, econoimcs.utoronto.ca/osborne (open
access, very nicely organized)
Microeconomic Analysis by Hal Varian, Chapters 26 and 27
(terse but useful)
Optimization Tools
4.3
The First Derivative
Let f : R !R. The derivative of f at x
be denoted Df (x
). Because
f (x) is a scalar function, we have:
Df (x
) =
df (x)
dx
The rst derivative can be used to approximate the value of f at
points close to x
. For small departures distances t , we have

f (x
+ t ) f (x
) + Df (x
)t .
Alternatively, we might write:
f (x) L(x|x
) f (x
) + Df (x
)(x x
)
where L(x|x
) denotes the linear approximation to f anchored at x
.
Optimization Tools
4.4
An Example of Linear Approximation
To illustrate how a linear approximation works, suppose that
f (x) = sin(x). We have Df (x) = cos(x). A local approximation to f (x)
is then
L(x|
x) = sin(
x) + cos(
x)(x
x)
Optimization Tools
4.5
Alternative Linear Approximations to sin(x)
Optimization Tools
4.6
Second Order Approximations
A second order Taylor series approximation can be employed when
the function to be approximated has continuous second derivatives.
We can dene a quadratic approximation to f (x) as:
Q(x; x
) = f (x
) + Df (x
)(x x
) +
1
2
(x x
)D
2
f (x
)(x x
)
The following gure illustrates the relationship between the underling
sine function and three different quadratic approximations.
Optimization Tools
4.7
Alternative Quadratic Approximations
Optimization Tools
4.8
The Gradient Vector
When f (x) is a scalar function with vector arguments, e.g. m = 1 or
f : R
n
!R, the gradient of f at x
is a vector whose coordinates are

the partial derivatives of f at x
:
D(f (x
)) =
@f (x
)
@x
1
, . . . ,
@f (x
)
@x
n
The gradient vector is also denoted rf (x
).
Optimization Tools
4.9
Denition
A quadratic form on R
n
is a real-valued function of the form:
Q(x
1
, . . . , x
n
) =
X
ij
a
ij
x
i
x
j
in which each term is monomial of degree two.
We can write this type of function compactly with vector-matrix
notation, i.e.
Q(x) = x
T
Ax
in which A is a symmetric matrix.
Optimization Tools
4.10
Quadratic Forms Two Dimensions
When n = 2, we have:
Q(x) = a
11
x
2
1
+ a
12
x
1
x
2
+ a
22
x
2
2
provided that
A =
a
11
1
2
a
12
1
2
a
12
a
22
The Jacobian matrix of a given function provides a typical symmetric

matrices which appears in quadratic forms.
Note that if A is a non-symmetric square matrix, the associated
quadratic form has the same value as the related symmetric matrix:
A
0
=
1
2
A +
1
2
A
T
Optimization Tools
4.11
Deniteness and Quadratic Forms
Recall our quadratic approximation to a function f :
f (x) f (x
) + Df (x
)(x x
) +
1
2
(x x
)
0
D
2
f (x
)(x x
))
Suppose that we have selected an x
such that Df (x
) = 0. Then the
value of f (x) is given by:
f (x) f (x
) + (x x
)
0
A(x x
))
where A =
1
2
D
2
f (x
).
If A is positive denite then x
is a local minimizer of f ().

If A is negative denite then x
is a local maximizer of f ().

Optimization Tools
4.12
Concavity
A function of one variable is concave if
f (tx + (1 t )y) tf (x) + (1 t )f (y)
For example, the sin(x) function is concave between x = 0.2 and
y = 1.6, as illustrated in the following gure.
Optimization Tools
4.13
Local Concavity of the Sine Function
Optimization Tools
4.14
Convexity
1 If f is a convex function, then f
00
(x) 0 for all x
2 If f is a convex function, then
f (x) f (y) + f
0
(y)|x y|
3 If f is a convex function, and f
0
(x
) = 0, then x
minimizes the
function f .
Optimization Tools
4.15
Unconstrained minimization
If f is differentiable at a local minimum x 2 U (open), then
rf (
x) = 0.
This is a necessary condition not a sufcient condition. (All local
minima satisfy this condition, but there exist points which are not local
minima which also satisfy this condition, e.g. local maxima or saddle
points.)
Optimization Tools
4.16
Descent directions
f : U !R differentiable

x 2 U (open)
If rf (
x)v < 0 then 9 2 R such that

f (
x + v) < f (
x) 8 2 (0, )
The vector v (above) is a descent direction at

x.
Recall that if rf (
x) 6= 0 then rf (
x) is the direction of steepest ascent

at

x.
This follows from the Cauchy-Swartz inequality
|x

y| = ||x||||y|| cos() ||x||||y||
Optimization Tools
4.17
Equality Constrained Optimization
minf (x)
subject to:
g(x) = 0 (P)
x 2 R
n
where
f and g are differentiable on R
n
.
g : R
n
!R
m
m n
Optimization Tools
4.18
Lagranges Theorem
Theorem
Lagrange If

x is a local minimum of (P), and the Jacobian matrix
rg(
x) has rank m, then there exist numbers

1
, . . . ,

m
such that
rf (
x) +
m
X
i =1
i
rg
i
(
x) = 0
The numbers

1
, . . . ,

m
are called Lagrange multipliers
The function L(x, ) = f (
x) +
P
m
i =1

i
g
i
(x) is the Lagrangian for (P).
Optimization Tools
4.19
Practical usefulness of Lagranges method
Solution of a constrained optimization problem with n variables and m
constraints can be equivalent to solving a nonlinear system of n + m
equations.
For economists, this result enormously simplies the formulation and
solution of market equilibrium models, because we are able to
incorporate multiple agents, each of which optimizes a separate
objective function subject to constraints.
Optimization Tools
4.20
Geometry of Constrained Optimization
Optimization Tools
4.21
Need for the regularity condition
The assumption that rank rg(
x) = m is a regularity condition.
Lagranges theorem is not valid unless the regularity condition holds.
EXAMPLE:
minx
1
(P) subject to
x
2
1
+ (x
2
1)
2
= 1
x
2
1
+ (x
2
+ 1)
2
= 1
Note: (P) has only one feasible point

x = (0, 0).
rf (
x) = (1, 0)
rg
1
(
x) = (0, 2)
rg
2
(
x) = (0, 2)
The Lagrange multipliers cannot exist here.
Optimization Tools
4.22
Irregular Example: No Multipliers Exist
5a
Consumer Choice Examples
Optimization and Human Behavior
Handout for Managerial Economics October 21, 2011
Thomas F. Rutherford, Center for Energy Policy and Economics, ETH Zrich
5a.1
A Choice Experiment
Thomas lives in Ann Arbor where he currently spends 30% of his income on rent. He has an
employment offer in Zrich which pays 50% more than he currently earns, but he is hesitant
to take the job because rental rates in Zrich are three times higher than in Ann Arbor. Assum-
ing that Thomas has CES preferences with elasticity of substitution ; on purely economic
grounds, should he move?
As is the case for all interesting questions in economics, the only good answer to this problem is It
depends.. 5a.2
Intuition
Thomass offer in Zrich does not pay him enough to live exactly the lifestyle that he enjoys in Ann
Arbor, as he would need a 60% raise to cover rent and consumption. The elasticity of substitution is key.
If it is high, he more willing substitutes consumption of goods and services for housing and thereby lowers
his cost of living in Zrich. On the other hand, if the elasticity is low, he is stuck in his ways, and the
move is a bad idea. 5a.3
Calibration to a Benchmark Equilibrium
We are given information about Thomass choices in Ann Arbor. This information is essentially an
observation of a benchmark equilibrium, consisting of the prevailing prices and quantities of goods demand.
The benchmark equilibrium data together with assumptions about elasticities are used to evaluate Thomass
choices after a discrete change in the economic environment. The steps involved in solving this little
textbook model are identical to those typically employed in applied general equilibrium analysis. 5a.4
Graphical Representation
5a.5
1
Preferences
The utility function:
U(C, H) = (C
+(1)H
)
1/
Exponent is dened by the elasticity of substitution, , as
= 11/.
The model of consumer choice is:
maxU(C, H) s.t. C+ p
H
H = M
5a.6
Demand
Derivation of demand functions which solve the utility maximization problem involves solving two
equations in two unknowns:
U/H
U/C
=
(1)H
1
C
1
= p
H
;
hence
H
C
=
1
p
H
Substituting into the budget constraint, we have:

H =
M
p
H
+
p
H
1
=
(1)
Mp
+(1)
p
1
H
and
C =
M
1+ p
H
1
p
H
+(1)
p
1
H
5a.7
Calibration
It is conventional in applied general equilibriumanalysis to employ exogenous elasticities and calibrated
value values. If we follow this approach, is then exogenous and is calibrated.
Choosing units so that the benchmark price of housing ( p
H
) is unity, we have:
= p
H

H/

M
Substitute into the demand function:
1+
H
=
1
;
and then solve for the preference parameter :
=
(1)
1/
1/
+(1)
1/
.
5a.8
Money Metric Utility
Substitute for in U(C, H), and denoting the base year expenditure on other goods as

C = (1)

M,
we have
U(C, H) =
(1)
1/
C
+
1/
H
1/
where the is a constant which may take on any positive value without altering the preference ordering.
It is convenient to assign this value to the benchmark expenditure, so that utility can be measured in money-
metric units at benchmark prices.
Noting that
1/
=
1
, we then can write the utility function as:
U(C, H) =

M
(1)
1/
5a.9
2
Indirect Utility
Formally, we have:
V(p
H
, M) =U(C(p
H
, M), M(p
H
, M)) =
M
+(1)
p
1
H
1/(1)
In money-metric terms, we can use benchmark income to normalize the utility function:
V(p
H
, M) =
M
(1 + p
1
H
)
1/(1)
5a.10
Demand Functions Calibrated Share Form
H =

H

V(p
H
, M)
p
U
p
H
C =

C

V(p
H
, M)
p
U
1
where
p
U
=
1 + p
1
H
1/(1)
5a.11
Should Thomas Move?
Thomass welfare level in Zrich can be easily computed in money-metric terms as:
V(p
H
= 3, M = 1.5) =
1.5
0.7+0.33
1
1/(1)
This expression cannot (to my knowledge) be solved in closed form, however it is easily to solve using
Excel. The critical value for is that which equates welfare in Zrich with welfare level in Ann Arbor, i.e.
V = 1. The numerical value is found to be
= 0.441. The general dependence of welfare on the and

can be illustrated in a contour diagram. 5a.12
Dependence of Welfare on Benchmark Shares and Elasticity
5a.13
Multivariable Optimization
The concept of multivariate optimization is important in managerial economics because many demand
and supply relations involve more than two variables. In demand analysis, it is typical to consider the
quantity sold as a function of the price of the product itself, the price of other goods, advertising, income,
and other factors. In cost analysis, cost is determined by output, input prices, the nature of technology, and
so on.. 5a.14
3
Optimal Advertising
To explore the concepts of multivariate optimization and the optimal level of advertising, consider
a hypothetical multivariate product demand function for CSI, Inc., where the demand for product Q is
determined by the price charged, P, and the level of advertising, A:
Q = 5, 00010P+40A+PA0.8A
2
0.5P
2
Determine the joint optimal price (P
) and level of advertising (A
) which maximize CSI output. 5a.15

First Order Conditions
Begin by calculating partial derivates of demand with respect to price and level of advertising:
Q
P
=10+AP
Q
A
= 40+P1.6A
First order conditions for maximization of demand are:
Q
P
= 0
Q
A
= 0
5a.16
Optimization = Solving Simultaneous Equations
Hence, the optimal level of price and advertising solve:
10+AP = 0
40+P1.6A = 0
Hence, P
= 40, A
= $5, 000 and the maximal output is Q
= 5, 800.
Note that in subsequent chapters we will learn that the policies which maximize output may differ from
those which maximal prot, depending on how production cost relates to output. 5a.17
Nonlinear Pricing
Consider a consumer choice model in which the two goods consist of telecommunication services (x)
and all other goods (y). Let the price of other goods is xed at unity. Telecommunication services are
somewhat special in that due to economies of scale, these are offered with potentially substantial quantity
discounts. Once a subscription fee of f CHF is made, services are offerred at a substantially reduced price.
In the absence of the connection fee, p
x
= 1. Telecommunication services made to customers who have
paid the connection fee are offered at a price of p
x
.
The consumer is assumed to have the following utility function:
maxU(x, y) = x
y
1
5a.18
A. Ignoring the subscription plan, solve for the quantity of telecommunication services demanded by
the consumer.
The standard consumer model is one of budget-constrained utility maximization. Hence, we solve
maxU(x, y) s.t. p
x
x +y = M. The rst order condition is:
U(x, y)/x
U(x, y)/y
=
p
x
1
Hence,
x
=
M
p
x
and
y
=
(1)M
1
5a.19
4
B. Assuming that the consumer chooses to buy a subscription. Show that she will buy the following
quantities:
x
=
M f
p
x
y
= (1)(M f )
If the consumer buys a subscription, the purchase quantity solves:
maxU(x, y)
s.t.
p
x
x +y = M f .
The rst order conditions are identical to the previous case, except that M is replaced by M f and
p
x
is replaced by p
x
.
5a.20
C. Holding p xed, what is the critical value of f such that the consumer is indifferent about buying a
subscription.
The critical value of f is that for which:
U(x
, y
) =U( x
, y
)
Substituting for U() we have:
M
p
x
(1)
M
1
1
=
M f
p
x
(1)
M f
1
1
Thus,
M
p
x
=
M f
p
x
and
f
= M
p
x
p
x
5a.21
D. Sketch the budget constraint and the optimal choice for a consumer who chooses not to accept the
subscription.
If a consumer buys a subscription, the maximum amount she can purchase of other goods is M f .
The slope of the budget line is p. If the optimal point on the subscription-based budget constraint
is associated with a lower indifference curve, then the consumer will not purchase a subscription:
5a.22
E. Holding f xed, graphically nd the maximum discount price level which would induce this con-
sumer to purchase additional units of telecommunication services ( p
x
).
Here we rotate the subscription based budget constraint around the y axis intercept to the point that it
is just tangent to the original indifference curve:
5
5a.23
F. Solve for p
h
( f ) analytically.
As shown above, the price which makes the consumer indifferent between taking a subscription or
not is:
p
( f ) =
1
f
M
1/
5a.24
G. Suppose that the marginal cost of supply for telecom services is 1. What combination of f and p
x
maximizes rm prot?
max
f
( f ) = f +(p( f ) 1) x = f +(p1)
M f
p( f )
The rst order for prot maximization is:
d
df
= 1
p

f
p
2
dp( f )
df
= 0
Applying the basic rules of calculus, we have:
dp( f )
df
=
1
1
f
M
1/1
1
M
=
p
1
M
Hence:
dp( f )
df
= 1 +

p
+
f
p
2
= 0 p = 1
5a.25
Basic idea: nonlinear pricing does not provide a means of increasing rm prots in the case of Cobb-
Douglas demand. The optimal subscription rate is zero ( f
= 0), and it is optimal to price at marginal cost

( p
= 1). 5a.26
Cobb-Douglas Calibration
Suzy consumes ice cream (x
1
) and soda (x
2
) for lunch every day, and she currently has one ice cream
and two sodas per week when they both cost 1 CHF. What Cobb-Douglas utility function is consistent with
Suzys choices over ice cream and soda. Write down demand functions which could extrapolate her optimal
choices to any expenditure (m) and prices (p
1
and p
2
). 5a.27
A Cobb-Douglas Calibration Exercise: Answer
Based current choices, we observe that Suzys budget shares for ice cream and sodas are 1/3 and 2/3,
respectively. The Cobb-Douglas utility function which describes her preferences is:
U(x
1
, x
2
) = x
1/3
1
x
2/3
2
and demand functions are
x
1
=
Y
3p
1
and
x
2
=
2Y
3p
2
5a.28
6
Calibration Exercise #2
Suppose that irregardless of relative prices, Suzy always has one soda before and one soda after eating
an ice cream. What utility function is consistent with these choices? Write down demand functions which
could extrapolate her optimal choices to any expenditure (m) and prices (p
1
and p
2
). 5a.29
Exercise # 2: Solution
Perfect complement preferences have the form:
U(x
1
, x
2
) = min(
x
1
a
1
,
x
2
a
2
)
in which the ratio
a
1
a
2
determines the ratio in which goods 1 and 2 are consumed. In the present example,
we have:
U(x
1
, x
2
) = min(x
1
,
x
2
2
)
and demand functions given by:
x
1
=
Y
p
1
+2p
2
and
x
2
= 2
Y
p
1
+2p
2
5a.30
Calibration Exercise #3
When Suzy gets to the lunch counter, she always asks about the price of ice cream and the price of soda.
If two sodas cost less than one ice cream, she has spends all of her money on soda. Otherwise she buys ice
cream. What utility function is consistent with these choices? Write down demand functions which could
extrapolate her optimal choices to any expenditure (m) and prices (p
1
and p
2
). 5a.31
Calibration Exercise #3 Solution
General perfect substitues preferences have the form:
U(x
1
, x
2
) = a
1
x
1
+a
2
x
2
in which the ratio
a
1
a
2
represents the marginal rate of substitution of good 1 for good 2. The demand functions
for these preferences are given by:
x
1
=
0 when
p
1
p
2
>
a
1
a
2
M
p
1
otherwise
x
2
=
0 when
p
1
p
2
<
a
1
a
2
M
p
2
otherwise
5a.32
7
Chapter 5: Production
Lecture Notes
Friday, October 21, 2011
Overview
I. Production Analysis
Total Product, Marginal Product, Average Product.
Isoquants.
Isocosts.
Cost Minimization
II. Cost Analysis
Total Cost, Variable Cost, Fixed Costs.
Cubic Cost Function.
Cost Relations.
III. Multi-Product Cost Functions
Production Analysis
Production Function
Q = F(K,L)
Q is quantity of output produced.
K is capital input.
L is labor input.
F is a functional form relating the inputs to output.
The maximum amount of output that can be
produced with K units of capital and L units of
labor.
Short-Run vs. Long-Run Decisions
Fixed vs. Variable Inputs
Production Function Algebraic Forms
Linear production function: inputs are perfect
substitutes.

Leontief production function: inputs are used in
fixed proportions.

Cobb-Douglas production function: inputs have a
degree of substitutability.
( )
b a
L K L K F Q = = ,
( ) bL aK L K F Q + = = ,
( ) { } cL bK L K F Q , min , = =
Productivity Measures:
Total Product
Total Product (TP): maximum output produced with
given amounts of inputs.
Example: Cobb-Douglas Production Function:
Q = F(K,L) = K
.5
L
.5
K is fixed at 16 units.
Short run Cobb-Douglass production function:
Q = (16)
.5
L
.5
= 4 L
.5
Total Product when 100 units of labor are used?
Q = 4 (100)
.5
= 4(10) = 40 units
Productivity Measures: Average Product of
an Input
Average Product of an Input: measure of output
produced per unit of input.
Average Product of Labor: AP
L
= Q/L.
Measures the output of an average worker.
Example: Q = F(K,L) = K
.5
L
.5

If the inputs are K = 16 and L = 16, then the average product of labor
is AP
L
= [(16)
0.5
(16)
0.5
]/16 = 1.
Average Product of Capital: AP
K
= Q/K.
Measures the output of an average unit of capital.
Example: Q = F(K,L) = K
.5
L
.5

If the inputs are K = 16 and L = 16, then the average product of
capital is AP
K
= [(16)
0.5
(16)
0.5
]/16 = 1.
Productivity Measures:
Marginal Product of an Input
Marginal Product on an Input: change in total
output attributable to the last unit of an input.
Marginal Product of Labor: MP
L
= AQ/AL
Measures the output produced by the last worker.
Slope of the short-run production function (with respect to
labor).
Marginal Product of Capital: MP
K
= AQ/AK
Measures the output produced by the last unit of capital.
When capital is allowed to vary in the short run, MP
K
is the
slope of the production function (with respect to capital).
Q
L
Q=F(K,L)
Increasing
Marginal
Returns
Diminishing
Marginal
Returns
Negative
Marginal
Returns
MP
AP
Increasing, Diminishing and Negative
Marginal Returns
Guiding the Production Process
Producing on the production function
Aligning incentives to induce maximum worker effort.
Employing the right level of inputs
When labor or capital vary in the short run, to
maximize profit a manager will hire:
labor until the value of marginal product of labor equals the
wage: VMP
L
= w, where VMP
L
= P x MP
L
.
capital until the value of marginal product of capital equals
the rental rate: VMP
K
= r, where VMP
K
= P x MP
K
.
Isoquant
Illustrates the long-run combinations of
inputs (K, L) that yield the producer the
same level of output.
The shape of an isoquant reflects the ease
with which a producer can substitute
among inputs while maintaining the same
level of output.
Marginal Rate of Technical Substitution
(MRTS)
The rate at which two inputs are
substituted while maintaining the same
output level.

K
L
KL
MP
MP
MRTS =
Linear Isoquants
Capital and labor are
perfect substitutes
Q = aK + bL
MRTS
KL
= b/a
Linear isoquants imply that
inputs are substituted at a
constant rate, independent
of the input levels
employed.

Q
3
Q
2
Q
1

Increasing
Output
L
K
Leontief Isoquants
Capital and labor are perfect
complements.
Capital and labor are used in
fixed-proportions.
Q = min {bK, cL}
Since capital and labor are
consumed in fixed
proportions there is no input
substitution along isoquants
(hence, no MRTS
KL
).
Q
3

Q
2

Q
1

K
Increasing
Output
L
Cobb-Douglas Isoquants
Inputs are not perfectly
substitutable.
Diminishing marginal rate
of technical substitution.
As less of one input is used in
the production process,
increasingly more of the other
input must be employed to
produce the same output
level.
Q = K
a
L
b
MRTS
KL
= MP
L
/MP
K
Q
1

Q
2
Q
3
K
L
Increasing
Output
Isocost
The combinations of inputs
that produce a given level of
output at the same cost:
wL + rK = C
Rearranging,
K= (1/r)C - (w/r)L
For given input prices, isocosts
farther from the origin are
associated with higher costs.
Changes in input prices change
the slope of the isocost line.
K
L
C
1

L
K
New Isocost Line for
a decrease in the
wage (price of labor:
w
0
> w
1
).
C
1
/r
C
1
/w
C
0

C
0
/w
C
0
/r
C/w
0
C/w
1
C/r
New Isocost Line
associated with higher
costs (C
0
< C
1
).
Cost Minimization
Marginal product per dollar spent should be
equal for all inputs:

But, this is just

r
w
MP
MP
r
MP
w
MP
K
L K L
= =
r
w
MRTS
KL
=
Cost Minimization
Q
L
K
Point of Cost
Minimization
Slope of Isocost
=
Slope of Isoquant
Optimal Input Substitution
A firm initially produces Q
0

by employing the
combination of inputs
represented by point A at a
cost of C
0
.
Suppose w
0
falls to w
1
.
The isocost curve rotates
counterclockwise; which
represents the same cost level
prior to the wage change.
To produce the same level of
output, Q
0
, the firm will
produce on a lower isocost line
(C
1
) at a point B.
The slope of the new isocost
line represents the lower wage
relative to the rental rate of
capital.
Q
0
0
A
L
K
C
0
/w
1 C
0
/w
0
C
1
/w
1
L
0
L
1
K
0
K
1
B
Cost Analysis
Types of Costs
Short-Run
Fixed costs (FC)
Sunk costs
Short-run variable
costs (VC)
Short-run total costs
(TC)
Long-Run
All costs are variable
No fixed costs
Total and Variable Costs
C(Q): Minimum total cost of
producing alternative levels
of output:

C(Q) = VC(Q) + FC

VC(Q): Costs that vary with
output.

FC: Costs that do not vary
with output.
$
Q
C(Q) = VC + FC
VC(Q)
FC
0
Fixed and Sunk Costs
FC: Costs that do not
change as output changes.

Sunk Cost: A cost that is
forever lost after it has been
paid.

Decision makers should
ignore sunk costs to
maximize profit or minimize
losses
$
Q
FC
C(Q) = VC + FC
VC(Q)
Some Definitions
Average Total Cost
ATC = AVC + AFC
ATC = C(Q)/Q

Average Variable Cost
AVC = VC(Q)/Q

Average Fixed Cost
AFC = FC/Q

Marginal Cost
MC = DC/DQ
$
Q
ATC
AVC
AFC
MC
MR
Fixed Cost
$
Q
ATC
AVC
MC
ATC
AVC
Q
0

AFC
Fixed Cost
Q
0
(ATC-AVC)

= Q
0
AFC

= Q
0
(FC/ Q
0
)

= FC
Variable Cost
$
Q
ATC
AVC
MC
AVC
Variable Cost
Q
0

Q
0
AVC

= Q
0
[VC(Q
0
)/ Q
0
]

= VC(Q
0
)

Minimum of AVC
$
Q
ATC
AVC
MC
ATC
Total Cost
Q
0

Q
0
ATC

= Q
0
[C(Q
0
)/ Q
0
]

= C(Q
0
)

Total Cost
Minimum of ATC
Cubic Cost Function
C(Q) = f + a Q + b Q
2
+ cQ
3

Marginal Cost?
MC(Q) = dC/dQ = a + 2bQ + 3cQ
2
An Example
Total Cost: C(Q) = 10 + Q + Q
2

Variable cost function:
VC(Q) = Q + Q
2
Variable cost of producing 2 units:
VC(2) = 2 + (2)
2
= 6
Fixed costs:
FC = 10
Marginal cost function:
MC(Q) = 1 + 2Q
Marginal cost of producing 2 units:
MC(2) = 1 + 2(2) = 5

Long-Run Average Costs
LRAC
$
Q
Economies
of Scale
Diseconomies
of Scale
Q*
Multi-Product Cost Function
C(Q
1
, Q
2
): Cost of jointly producing two
outputs.
General function form:
( )
2
2
2
1 2 1 2 1
, cQ bQ Q aQ f Q Q C + + + =
Economies of Scope
C(Q
1
, 0) + C(0, Q
2
) > C(Q
1
, Q
2
).
It is cheaper to produce the two outputs jointly
instead of separately.
Example:
It is cheaper for Time-Warner to produce Internet
connections and Instant Messaging services jointly
than separately.
Cost Complementarity
The marginal cost of producing good 1
declines as more of good two is produced:

AMC
1
(Q
1
,Q
2
)

/AQ
2
< 0.

Example:
Cow hides and steaks.
Quadratic Multi-Product Cost Function
C(Q
1
, Q
2
) = f + aQ
1
Q
2
+ (Q
1
)
2
+ (Q
2
)
2

MC
1
(Q
1
, Q
2
) = aQ
2
+ 2Q
1

MC
2
(Q
1
, Q
2
) = aQ
1
+ 2Q
2

Cost complementarity: a < 0
Economies of scope: f > aQ
1
Q
2
C(Q
1
,0) + C(0, Q
2
) = f + (Q
1
)
2
+ f + (Q
2
)
2

C(Q
1
, Q
2
) = f + aQ
1
Q
2
+ (Q
1
)
2
+ (Q
2
)
2

f > aQ
1
Q
2
: Joint production is cheaper

A Numerical Example:
C(Q
1
, Q
2
) = 90 - 2Q
1
Q
2
+ (Q
1
)
2
+ (Q
2
)
2

Cost Complementarity?
Yes, since a = -2 < 0
MC
1
(Q
1
, Q
2
) = -2Q
2
+ 2Q
1

Economies of Scope?
Yes, since 90 > -2Q
1
Q
2

Conclusion
To maximize profits (minimize costs) managers
must use inputs such that the value of marginal of
each input reflects price the firm must pay to
employ the input.
The optimal mix of inputs is achieved when the
MRTS
KL
= (w/r).
Cost functions are the foundation for helping to
determine profit-maximizing behavior in future
chapters.

Managerial Economics

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Managerial Economics

Hochgeladen von

Copyright:

Verfügbare Formate

The Nature and Scope

and the equilibrium price is:

and the equilibrium price is:

and the equilibrium price is:

and the equilibrium price is:

= c), hence price is determined by the inverse

, the equilibrium price without taxes, and

, the equilibrium price without taxes, and

, the equilibrium price without taxes, and

, the equilibrium price without taxes, and

which maximizes T. For t > t

, tax revenue decreases

which maximizes T. For t > t

, tax revenue decreases

which maximizes T. For t > t

, tax revenue decreases

denote the quantity for which

# 0, and when > 1, Q

(df 1,df 2,)

(df 1,df 2,0.05)

. We then construct an interval around

at a signicance level of . Such tests simply

at the signicance level

against the alternative

at the signicance level (1-tail t-test),

we reject the null

. For small departures distances t , we have

) denotes the linear approximation to f anchored at x

is a vector whose coordinates are

The gradient vector is also denoted rf (x

The Jacobian matrix of a given function provides a typical symmetric

is a local minimizer of f ().

is a local maximizer of f ().

x)v < 0 then 9 2 R such that

x) is the direction of steepest ascent

x) has rank m, then there exist numbers

Substituting into the budget constraint, we have:

V = 1. The numerical value is found to be

= 0.441. The general dependence of welfare on the and

) and level of advertising (A

) which maximize CSI output. 5a.15

= $5, 000 and the maximal output is Q

= 0), and it is optimal to price at marginal cost

Das könnte Ihnen auch gefallen