Public Goods

Rationales for Public

Policy: Market Failures

,' T h e idealized competitive model produces a Pareto-efficient allocation o f goods.

'

That is, the utility-maximizing behavior o f persons and the profit-mqimizing behavior o f firms will, through the "invisible hand," distribute goods in such a way that no
one could be better-off without making anyone else worse-off. Pareto efficiency
thus arises through voluntary acttons without any need for public policy. Economic
reality, however, rarely corresponds perfectly t o the assumptions o f the idealized
competitive model. In the following sections, w e discuss violations o f the assumptions that underlie the competitive model. These violations constitute market failures; that is, situations in which decentralized behavior does not lead t o Pareto efficiency. Traditional market failures are shown as circumstances in which social
surplus is larger under some alternative allocation t o that resulting in the market
equilibrium. Public goods, externalities, natural monopolies, and information asymmetries are the four commonly-recognized market failures. They provide the tradiin private affairs.
tional economic rationales for public pal-t~cipation

PUBLIC GOODS
T h e term public, or collectlve, goods appears frequently in the literature o f policy
analysis and economics. T h e blanket use o f the term, however, obscures important
differences among the variety o f public goods in terms o f the nature o f the market
failure, and, consequently, the appropriate public policy response. W e begin with a
basic question that should be raised when considering any market failure: W h y doesn't
the market allocate this particular good efficiently? The simplest approach t o providing an answer involves contrasting public goods with private goods.

75

T w o primary characteristics define private goods: rivalry in consumption and

excludability in ownership and use. Rivalrous consumption means that what one consumes cannot be consumed by another; a perfectly private good is characterized b y
complete rivalry in consumption. Excludable ownership means that one has control
over use o f the good; a perfectly private good is characterized by complete excludability. For example, shoes are private goods because when one wears them, no one
else can (rivalrous consumption) and, because when one owns them, one can determine w h o gets t o wear them at any particular time (excludable ownership).
Public goods, on the other hand, are, in varying degrees, nonrivalrous in consumption, nonexcludable in use, or both. In other words, w e consider any good that is
not purely private to be a public good. A good is nonrivalrous in consumption when
more than one person can derive consumption benefits from a given level o f supply
at the same time. For example, a given level o f national defense is nonrivalrous in
consumption because all citizens benefit from i t without reducing the benefits o f
others-a new citizen enjoys benefits without reducing the benefits o f those already
being defended. (Each person, however, may value the commonly provided level o f
defense differently.) A good is nonexcludable i f it is impractical for one person t o
maintain exclusive control over i t s use. For example, species offish that range widely
in the ocean are usually nonexcludable in use because they move freely among regions such that no individual can effectively exclude others from harvesting them.
In practice, a third characteristic related to demand, congestibility, is useful in
describing public goods. Levels o f demand for a good often determine the extent to
which markets supply publ~cgoods at inefficient levels. A good is congested if the
marginal social cost o f consumption exceeds the marginal private cost o f consumption. For example, a few people may be able t o hike in a wilderness area without interfering with each other's enjoyment o f the experience (a case o f low demand) so
that the marginal social cost o f consumption equals the marginal private cost o f consumption and there is no congestion. A larger number o f hikers may reduce each
other's enjoyment o f the wilderness experience (a case o f high demand) such that
the marginal social cost o f consumption exceeds the marginal private cost o f consumption and, therefore, the wilderness is congested. Later w e define and interpret
the divergence between social and private marginal costs as an externality.

Excludability and Properly Rights

Excludability implies that some individual can exclude others from use o f the
good. In most public policy contexts in developed democracies, power t o exclude
others from use o f a good is dependent on property rights granted and enforced by
the state and its (judicial) organs. Property rights are relationships among people concerning the use o f things.' These relationships involve claims by rights holders that

'For a seminal discussion o f property rights, see Eirik Furubotn and Svetozar Pejovich. "Property

Rights and Economic Theory: A Survey o f Recent Literature," Journal of EconomIc Llterature. Vol. 10.
no. 4. 1972, pp. 1137-62. For a current review o f ~ r o p e r t yright issues, see David L. Weimer, ed., The POIItIcal Economy of Property Rights: InstitutIonal Change and Credrb111t~
In the Reform of Centrally Planned

Economies (New York: Cambridge University Press. 1997), pp. 1-19.

Public Goods

Rationales for Public

Policy: Market Failures

.
-

'

,' The ideal~zedcompetitive model produces a Pareto-efficient allocation o f goods.

' That is, the utility-maximizing behavior o f persons and the profit-mqimizing behavior o f firms will, through the "invisible hand," distribute goods in such a way that no
one could be better-off without making anyone else worse-off. Pareto efficiency
thus arises through voluntary actions without any need for public policy. Economic
reality, however, rarely corresponds perFectly t o the assumptions o f the idealized
competitive model. In the following sections, w e discuss violations o f the assumptions that underlie the competitive model. These violations constitute market failures; that is, situations in which decentralized behavior does not lead t o Pareto e f i ciency. Traditional market failures are shown as circumstances in which social
surplus is larger under some alternative allocation t o that resulting in the market
equilibrium. Public goods, externalities, natural monopolies, and information asymmetries are the four commonly-recognized market failures. They provide the traditional economic rationales for public participation in private affairs.

PUBLIC GOODS
The term public, or collective, goods appears frequently in the literature o f policy
analysis and economics. T h e blanket use o f the term, however, obscures important
differences among the variety o f public goods in terms o f the nature o f the market
failure, and, consequently, the appropriate public policy response. W e begin with a
basic question that should be raised when considering any market failure: Why doesn't
the market allocate this particular good efficiently? The simplest approach t o providing an answer involves contrasting public goods w i t h private goods.

75

T w o primary characteristics define private goods: rivalry in consumption and

excludability in ownership and use. Rivalrous consumption means that what one consumes cannot be consumed by another; a perfectly private good is characterized by
complete rivalry in consumption. Excludable ownershb means that one has control
over use o f the good; a perfectly private good is characterized by complete excludability. For example, shoes are private goods because when one wears them, no one
else can (rivalrous consumption) and, because when one owns them, one can determine who gets t o wear them at any particular time (excludable ownership).
Public goods, on the other hand, are, in varying degrees, nonrivalrous in consumption, nonexcludable in use, or both. In other words, w e consider any good that is
not purely private to be a public good. A good is nonrivalrous in consumption when
more than one person can derive consumption benefits from a given level o f supply
at the same time. For example, a given level o f national defense is nonrivalrous in
consumption because all citizens benefit from i t without reducing the benefits o f
others-a new citizen enjoys benefits without reducing the benefits o f those already
being defended. (Each person, however, may value the commonly provided level o f
defense differently.) A good is nonexcludable if i t is impractical for one person to
maintain exclusive control over i t s use. For example, species o f fish that range widely
in the ocean are usually nonexcludable in use because they move freely among regions such that no individual can effectively exclude others from harvesting them.
In practice, a third characteristic related t o demand, congestibility, is useful in
describing public goods. Levels o f demand for a good often determine the extent t o
which markets supply publlc goods at inefficient levels. A good is congested i f the
marginal social cost o f consumption exceeds the marginal private cost o f consumption. For example, a few people may be able t o hike in a wilderness area without interfering with each other's enjoyment o f the experience (a case o f low demand) so
that the marginal social cost of consumption equals the marginal private cost o f consumption and there is no congestion. A larger number o f hikers may reduce each
other's enjoyment o f the wilderness experience (a case o f high demand) such that
the marginal social cost o f consumption exceeds the marginal private cost o f consumption and, therefore, the wilderness is congested. Later w e define and interpret
the divergence between social and private marginal costs as an externality.

Excludability and Property Rights

Excludability implies that some individual can exclude others from use o f the
good. In most public policy contexts in developed democracies, power t o exclude
others from use of a good is dependent on property rights granted and enforced by
the state and its (judicial) organs. Property rights are relationships among people concerning the use o f things.' These relationships involve claims by rights holders that

'For a sern~naldlscuss~onof property r~ghts,see Elr~kFurubotn and Svetozar Pe,ovlch, Property

R~ghtsand Econorn~cTheory A Survey of Recent Llterature," Journal of Econornh L~terature.Vol 10.
no 4, 1972, pp 1137-62 For a current revlew o f property r~ght
Issues, see Dav~d
L We~rner,ed , The Po11t1calEconomy of Property R~ghts lnstrtutronal Change and Credrbrl~t~
ln the Reform of Centrally Planned
Economres (New York Carnbrldge Un~vers~ty
Press. 1997), pp 1-19

..
76

Rationales for Public Policy: Market Failures

Chap. 5

impose dut~eson others. In contrast to political systems in which institutional

arrangements create and enforce property rights, there are anarchic, or Hobbesian,
systems with no government and no constraining norms or conventions in which
physical force alone determines the power to exclude.
Property rights can be partitioned in complex ways beyond what we think o f
as simple ownership.' For example, a farmer may have a property right that allows
using water from a river only during specific months o f the year. However, for the
purposes o f a discussion o f excludability, we treat goods as being controlled or
"owned" by a single actor. Effective property rights are characterized by clear and
complete allocation o f claims and high levels of compliance by those who owe the
corresponding duties. In this context. where the claim is to exclusive use o f the
good, compliance simply means accepting exclusion. De lure property rights, which
are granted by the state, are typically clear though sometimes incomplete. These de
jure property rights, however, may be attenuated, or in some cases superseded, by
extralegal behaviors such as trespass, squatting, poaching, or custom. Behaviors
such as these may give rise to de facto property rights, the claims that actually receive compliance from duty bearers. Sometimes de lure property rights do not exist
because changes in technology or relative prices create new goods that fall outside
o f existing allocations. For example, advances in medical technology have made fetal
tissue valuable, but the right to its use is yet unclear. Note that de facto property
rights may or may not present excludability problems, though often they involve subal
systems, vigistantial costs to the claimants if they must employ ~ h ~ s i cprotection
lance, stealth, or retaliation to enforce them. I f these costs become too high, then individuals may abandon attempts to exclude others from use o f th'e good. In these
cases, the good is effectively nonexcludable.
Nonrivalrous Goods

As we concluded in our discussion o f efficient pricing, the production o f a private good, involving rivalrous consumption, will be in equilibrium at a level where
price equals marginal cost (P = MC). O n the demand side, the marginal benefits consumers receive from additional consumption must equal price (MB = P) in equilibrium. Therefore, marginal benefits equal marginal cost (MB = MC). Essentially the
same principle applies to nonrivalrous goods. Because all consumers receive marginal
benefits from the additional unit o f the nonrivalrous good, however, i t should be produced if the sum o f all individual consumers' marginal benefits exceeds the marginal
cost of producing it. Only when output is increased to the point where the sum o f
the marginal benefits equals the marginal cost o f production is the quantity produced
efficient.
The sum of marginal benefits for any level o f output o f a purely nonrivalrous
pubiic good (whether excludable or not) is obtained by vertically summing the marginal benefit schedules (demand schedules) o f all individuals at that output level.' Re-

'yoram Barzel. Econornrc Anolys,s of Property R~ghts(New York: Cambridge University Press.

1989).
'Paul A. Samuelson. "Diagrammatic Exposition o f a Theory o f Public Expenditure," Review of
37. no. 4, 1955. pp. 350-56.

Econorn,cs ondStotistrcs. Vol.

Public Goods

77

peating the process at each output level yields the entire social, or aggregate, marginal benefit schedule. This contrasts with the derivation o f the social marginal benefit schedule for a private good-individual
marginal benefit schedules (demand
schedules) are added horizontally because each unit produced can only be consumed
by one oerson.
Figure 5.1 illustrates the different approaches. Panel (a) represents the demand for a rivalrous good and panel (b) the demand for a nonrivalrous good. In both
cases, the demand schedules o f the individual consumers, Dilbert and Dogbert, for
the good appear as Dl and D2,,respectively. The market demand for the rivalrous
good results from the horizontal addition o f individual demands Dl and D2. For example, at price po. Dilbert demands Q, and Dogbert demands Q2 so that the total
quantity demanded i s Qo (equal to Q, + Q2). Price Poalso happens to be the price
that equates marginal social benefits with marginal social costs (given by supply
schedule 3).
Panel (b) presents the parallel situation for the nonrivalrous good, with the
caveat that we are unlikely to be able to observe individual demand schedules for
public goods (for reasons we discuss below) Here, the social marginal benefit at Q,
(MBJ is the sum o f the marginal benefits enjoyed by Dilbert (MBI) and Dogbert
(MB2). Given the indicated supply schedule (S), Qs equates marginal social benefit
with marginal social cost and therefore corresponds to the Pareto-efficient output
level of the nonrivalrous good.
Notice that in this example, the upward-sloping supply schedule indicates that
higher output levels have higher marginal costs (a brighter streetlight costs more to
operate than a dimmer one); it is only the marg~nalcost of consumption that is equal,
and zero, for each person (each person can each look at what is illuminated by the
light without interfering with the other). In other words, there are zero marginal social costs of consumption but positive marginal costs o f production at each level o f
supply.

A crucial distinction between rivalrous and nonrivalrous goods is that the valuations of individual consumers cannot directly tell us how much o f the nonrivalrous
good should be provided-only the sum o f the valuations can tell us that. Once an
output level has been chosen, every person must consume it. Therefore, the various
values different persons place on the chosen output level are not revealed by their
purchases as they would be in a market for a rivalrous good. Thus, price neither
serves as an allocative mechanism nor reveals marginal benefits as it does for a rivalrous good.
Why wouldn't individuals supply the level o f output that equates marginal cost
with the sum o f individual marginal benefits as they would in a market for a rivalrous
good? Returning to panel (b) o f Figure 5.1, if Dogbert, who has demand D2, for the
conrivalrous good makes his own decision, he will purchase a quantity Q2, where his
own marginal benefit schedule crosses the supply schedule, which is less than the socially optimal quantity, Qs. Dilbert, who has demand (D,), would purchase an even
smaller quantity, Q,,if he were the only demander in the market. But ifDilbert knew
that Dogbert would make a purchase, he would not find it in his own self-interest to
purchase any of the nonrivalrous good because, as the marginal benefit schedules are
drawn, Dogbert's purchase quantity would exceed the amount that Dilbert would
want to purchase on his own. H e would have an incentive tofree ride on Dogbert's
consumption, which he gets to enjoy because o f nonrivalry. In other words, once

Rationales for Public Policy: Market Failures

Chap. 5

impose duties on others. In contrast to political systems in which institutional

arrangements create and enforce property rights, there are anarchic, or Hobbesian,
systems with no government and no constraining norms or conventions in which
physical force alone determines the power to exclude.
Property rights can be partitioned in complex ways beyond what we think of
~ . ~example, a farmer may have a property right that allows
as simple o w n e r s h ~For
using water from a river only during specific months o f the year. However, for the
purposes o f a discussion o f excludability, we treat goods as being controlled or
"owned" by a single actor. Effective property rights are characterized by clear and
complete allocation o f claims and high levels o f compliance by those who owe the
corresponding duties. In this context, where the claim is to exclusive use o f the
good, compliance simply means accepting exclusion. D e ~ u r eproperty rights, which
These de
are granted by the state, are typically clear though sometimes~incomplete.
jure property rights, however, may be attenuated, or in some cases superseded, by
extralegal behaviors such as trespass, squatting, poaching, or custom. Behaviors
such as these may give rise to de focto property rights, the claims that actually receive compliance from duty bearers. Sometimes de ~ureproperty rights do not exist
because changes in technology or relative prices create new goods that fall outside
of existing allocations. For example, advances in medical technology have made fetal
tissue valuable, but the right to its use is yet unclear. Note that deJacto property
rights may or may not present excludability problems, though often they involve substantial costs to the claimants if they must employ physical protection systems, vigilance, stealth, or retaliation to enforce them. If these costs become too high, then individuals may abandon attempts to exclude others from use o f th'e good. In these
cases, the good IS effectively nonexcludable.

Nonrivalrous Goods
As we concluded in our discussion o f efficient pricing, the production o f a private good, involving rivalrous consumption, will be in equilibrium at a level where
price equals marginal cost (P = MC). O n the demand side, the marginalbenefits consumers receive from additional consumption must equal price (MB = P) in equilibrium. Therefore, marginal benefits equal marginal cost (MB = MC). Essentially the
same principle applies to nonrivalrous goods. Because all consumers receive marginal
benefits from the additional unit o f the nonrivalrous good, however, it should be produced i f the sum o f all individual consumers' marginal benefits exceeds the marginal
cost of producing it. Only when output is increased to the point where the sum of
the marginal benefits equals the marginal cost o f production i s the quantity produced
eficient.
The sum o f marginal benefits for any level o f output o f a purely nonrivalrous
public good (whether excludable or not) i s obtained by vertically summing the marginal benefit schedules (demand schedules) o f all individuals at that output level3 Re-

'yoram Barzel. Econornrc Analysis of Property Rights (New York: Cambridge University Press.

1989).
'Paul A. Samuelson. "Diagrammatic Exposition of a Theory of Public Expenditure," Review of
1955, pp. 35G56.

Economics ondStotktrcs, Vol. 37, no. 4,

Public Goods

77

peating the process at each output level yields the entire social, or aggregate, marginal benefit schedule. This contrasts with the derivation o f the social marginal benefit schedule for a private good-individual
marginal benefit schedules (demand
schedules) are added horizontally because each unit produced can only be consumed
by one person.
Figure 5.1 illustrates the different approaches. Panel (a) represents the demand for a rivalrous good and panel (b) the demand for a nonrivalrous good. In both
cases, the demand schedules of the individual consumers, Dilbert and Dogbert, for
the good appear as D, and D2,,respectively. The market demand for the rivalrous
good results from the horizontal addition o f individual demands Dl and D2.For example, at price Po, Dllbert demands Q, and Dogbert demands Q2 so that the total
quantity demanded is Qo (equal to Q, + Q2). Price Poalso happens to be the price
that equates marginal social benefits with marginal social costs (given by supply
schedule S).
Panel (b) presents the parallel situation for the nonrivalrous good, with the
caveat that we are unlikely to be able to observe individual demand schedules for
public goods (for reasons we discuss below). Here, the social marginal benefit at QS
(MBJ) is the sum o f the marginal benefits enjoyed by Dilbert (MB,) and Dogbert
(MB2). Given the indicated supply schedule (S), QS equates marginal social benefit
with marginal social cost and therefore corresponds to the Pareto-efficient output
level o f the nonrivalrous good.
Notice that rn this example, the upward-sloping supply schedule indicates that
higher output levels have higher marginal costs (a brighter streetlight costs more to
operate than a dimmer one); it i s only the marginal cost o f consumption that is equal,
and zero, for each person (each person can each look at what is illuminated by the
light without interfering with the other). In other words, there are zero marginal social costs o f consumption but positive marginal costs of production at each level o f
supply.
A cruciai distinction between rivalrous and nonrivalrous goods is that the valuations o f individual consumers cannot directly tell us how much of the nonrivalrous
good should be prov~ded--onlythe sum of the valuations can tell us that. Once an
output level has been chosen, every person must consume it. Therefore, the various
values different persons place on the chosen output level are not revealed by their
purchases as they would be in a market for a rivalrous good. Thus, price neither
serves as an allocative mechanism nor reveals marginal benefits as it does for a rrvalrous good.
Why wouldn't individuals supply the level o f output that equates marginal cost
with the sum of individual marginal benefits as they would in a market for a rivalrous
good? Returning to panel (b) o f Figure 5.1, if Dogbert, who has demand D2,
for the
rionrivalrous good makes his own decision, he will purchase a quantity Q2, where his
own marglnal benefit schedule crosses the supply schedule, which is less than the socially optimal quantity, Q5. Dilbert, who has demand (Dl), would purchase an even
smaller quantity, Q,,if he were the only demander in the market. But if Dilbert knew
that Dogbert would make a purchase, he would not find i t in his own self-interest to
purchase any of the nonrivalrous good because, as the marginal benefit schedules are
drawn, Dogbert's purchase quantity would exceed the amount that Dilbert would
want to purchase on his own. He would have an incentive tofree ride on Dogbert's
consumption, which he gets to enjoy because o f nonrivalry. In other words, once

Rationales for Public Policy: Market Failures

Chap. 5

Horizontal summation
of demand schedules

&=QI

+Q2

Public Goods

79

Dogbert purchases Q2units, Dilbert would not find it in his personal interest t o purchase any additional units on his o w n because each unit he purchases gives him less
individual benefit than its price.
A market could still generate the optimal amount o f the nonrivalrous good if all
consumers would honestly reveal their marginal valuations. (Notice that one o f the
great advantages o f the market for rivalrous goods is that consumers automatically
reveal their marginal valuations with their purchases). Consumers do not normally
have an incentive t o reveal honestly their marginal valuations, however, when they
cannot be excluded from consumption o f the good.

Congestibility:The Role of Demand

Quantitymime
(a) Rivalrous Demand

Vertical summation of
marginal benefit schedules:
MB, = M B I + MB2

Quantitymime
(b) Nonrivalrous Good

Figure 5.1 Demand Summation for Rivalrous and Nonrivalrous

Goods

An economically relevant classification o f public goods requires attention to

more than their physical characteristics. At some level of demand, consumption o f a
good by one person may raise the marginal costs other persons face in consuming the
good, so that the marginal social cost o f consumption exceeds the marginal private
cost. Later, w e will define the divergence between social and private cost as an externality. In the context o f congestibility, we are dealing with a particular type o f externality-an externality of consumption inflicted only on other consumers o f the good.
Whether or not a particular good is congested at any particular time depends
on the level o f demand for the good at that time. Changes in technology, population,
income, or relative prices can shift demand from levels that do not involve externalities o f consumption to levels that do. For example, a road that could accommodate
one thousand vehicles per day without traffic delays might involve substantial traffic
delays in accommodating t w o thousand vehicles. In some cases, seasonal, or even
daily, shifts in demand may change the externalities o f consumption, making a good
more or less congested. For example, drivers may face substantial traffic delays during rush hour, but no delays during midday.
W e must be careful t o distinguish between the marginal social cost o f consumption and the marginal cost o f production. A purely nonrivalrous and noncongestible good exhibits zero marginal costs o f consumption. Yet, increments o f the
good (unless they occur naturally) require various factor inputs t o produce. For instance, one way t o increase the level o f defense is t o increase readiness, say by
shooting o f f more ammunition in practice. But it takes labor, machines, and materials
to produce the ammunition, things that could be used t o produce other goods instead. Thus, the marginal cost o f production ofdefense is not zero; the marginal cost
o f consumption is likely t o be zero, however, for a given level o f supply.
Some care is required in thinking about congestion in cases in which supply
cannot be added in arbitrarily small units. Many goods that display nonrivalry in consumption also display "lumpiness" in supply. For example, one cannot sirnnly nJd
small units o f physical capacity to an existrng bridge. T o provide additionai units of
capacity, one must typically either build a new bridge or "double-deck" the existing
one. Either approach provides a large addition t o capacity. But this "lumpines:" 15 i i ~ relevant t o the determination o f congestion. T h e important consideration is \/..~he'-lher
cr not the external costs o f consumptior; 3re positive at the available level o f supply.
T o summarize, three characteristics determine the specific nature o f the pub!ic
good (and hence the nature o f the inefficiency that would result from market sup-

Rationales for Public Policy: Market Failures

Chap. 5

Public Goods

79

Dogbert purchases Q2units, Dilbert would not find i t in his personal interest t o purchase any additional units on his own because each unit he purchases gives him less
individual benefit than its price.
A market could still generate the optimal amount o f the nonrivalrous good i f all #,
consumers would honestly reveal their marginal valuations. (Notice that one o f the
great advantages o f the market for rivalrous goods is that consumers automatically
reveal their marginal valuations with their purchases). Consumers do not normally
have an incentive to reveal honestly their marginal valuations, however, when they
cannot be excluded from consumption o f the good.

Horizontal summation
of demand schedules
Q o = Q l + Q2

Congestibility: The Role of Demand

Quantitymime
(a) Rivalrous Demand

Vertical summation of
marginal benefit schedules:
MB, = M B I + MB2

Quantitymime
(b) Nonrivalrous Good

Figure 5.1 D e m a n d Summation for Rivalrous and Nonrivalrous

Gcods

An economically relevant classification o f public goods requires attention to

more than their physical characteristics. At some level o f demand, consumption o f a
good by one person may raise the marginal costs other persons face in consuming the
good, so that the marginal social cost o f consumption exceeds the marginal private
cost. Later, we will define the divergence between social and private cost as an externality. In the context o f congestibility, we are dealing with a particular type o f externality-an externality of consumption inflicted only on other consumers o f the good.
Whether or not a particular good is congested at any particular time depends
on the level o f demand for the good at that time. Changes in technology, population,
income, or relative prices can shift demand from levels that do not involve externalities o f consumption to levels that do. For example, a road that could accommodate
one thousand vehicles per day without traffic delays might involve substantial traffic
delays in accommodating t w o thousand vehicles. In some cases, seasonal, or even
daily, shifts in demand may change the externalities o f consumption, making a good
more or less congested. For example, drivers may face substantial traffic delays during rush hour, but no delays during midday.
W e must be careful to distinguish between the marginal social cost of consumption and the marginal cost o f production. A purely nonrivalrous and noncongestible good exhibits zero marginal costs o f consumption. Yet, increments o f the
good (unless they occur naturally) require various factor inputs t o produce. For instance, one way t o increase the level o f defense is to increase readiness, say by
shooting off more ammunition in practice. But i t takes labor, machines, and materials
to produce the ammunition, things that could be used to produce other goods instead. Thus, the marginal cost o f production o f defense is not zero; the marginal cost
o f consumption is likely to be zero, however, for a glven level o f supply.
Some care is required in thinking about congestion in cases in which supply
cannot be added in arbitrarily small units. Many goods that display nonrivalry in consurnpt~onalso display "lumpiness" in supply. For example, one cannot simnly a3d
small umts o f physrcal capacity to an exrstlng bridge. T o provide additionai units of'
capacity, one must typically either build a new bridge or "double-decY the exrsting
one. Either approach provides a large addition to capacity. But this "lurnpines:" 15 ii-relevant to the determination o f congestion. The important consideration i-.\/%.~he'-her
GT not the external costs o f consumptior, are positive at the availcble level of' supply.
determine the specific natul-e o f the pub!ic
T o summarize, three characterist,~~
good (and hence the nature o f the inefficiency that would rcsult from marltet sup-

Chap. 5

Rationales for Public Policy: Market Failures

Public Goods

ply): the degree o f rivalry in consumption; the extent o f excludability, or exclusiveness, in use; and the existence o f congestion. The presence o f nonrivalry, nonexcludability, or congestion arising from changes in levels o f demand can lead to the failure
o f markets t o achieve Pareto efficiency. The presence o f either nonrivalry or nonexcludability is a necessary condition for the existence o f a public good market failure.

A Classification of Public Goods

Figure 5.2 presents the basic taxonomy o f public goods with the rivalrous/
nonrivalrous distinction labeling the columns and excludability/nonexcludabilitydistinction labellng the rows. Additionally, the diagonals within the cells separate congested from uncongested cases. By the way, note that the definitions o f public goods
that we provide differ starkly from a common usage o f the term "public good' to describe any good provided by government. In fact, governments, and indeed markets.
prov~deboth public and private goods as defined.

NONRIVALROUS

RIVALROUS

I
5m

D
3
A

?5

NW1

,/

1
'

Uncongested: Private Good

/'
/

NW2

NEI

./
/

Uncongested: Toll Good

No private supply at
//
,/
Congested:
efficient price of zero; /
underconsumption / / / Toll Good with
1
at any positive 1/ / '
Crowding
//
private Good wit/? price.
/
/ / / ~ o ~ ~ s ~ t nErt~n7zfllih~
~..-~. .r.-.
ion
/
/
/
private suvvlv. can be I
' I
/
/
/
efficient if price at marginal
Overconsumption because
/
social cost; peak-load pricing
,/
consumers respond to price
/
required if congestion variable.
/
rather
than
marginal
social
cost
/
SEI /
SW1 ,//
/
Uncongested: "Free Good"
/
Unconeested: Pure Public G o o d , / / c _

Efficient Market Supply

- -

.//

- r---

(efficiency until
,//
Ope11Access possible; some
Congested:
demand grbws to
and Cornmon private supply in / /
U exceed supply / / / PI-opert)'Resollrces privileged and / / ' ~ ~ n b i e nPublic
t
Good
with Consumption
at zero price.,,'
c~~~~~~~~
respond to intermediate//
z
groupS. / I /
Externality
1' marginal private cost rather
0
/
/
z
than marginal social cost- .
///
Overconsumption because
//overconsumption. rent dissipation,
consumers ignore external cost.
//land underinvestment in presentation. / /

5
%

Figure 5.2 A Classification of Goods: Private and Public

81

Rivalry, Excludability: Private Goods. The northwest (NW) cell defines private goods, characterized by both rivalry in consumption and excludability in
use: shoes, books, bread, and the other things we commonly purchase and own. In
the absence o f congestion or other market failures, the self-interested actions o f consumers and firms elicit and allocate these goods efficiently so that government intervention would have to be justified by some other rationale than the promotion o f efficiencv.
When the private good is congested (that is, i t exhibits externalities o f consumption), market supply is generally not efficient. (In competitive markets the marginal social costs o f the good equals not just the price-the marginal cost o f production-but the sum o f price and the marginal social costs o f consumption.) Rather
than consider such situations as public goods market failures, it is more common to
treat them under the general heading o f externality market failures. Consequently,
we postpone consideration o f category N W 2 in Figure 5.2 to our discussion o f externalities.
Nonrivalry, Excludability: Toll Goods. The northeast cell (NE) includes
those goods characterized by nonrivalry in consumption and excludability. Such
goods are often referred to as toll goods. Prominent examples include bridges and
roads that, once built, can carry significant levels o f traffic without crowding. Other
examples include goods that occur naturally such as wilderness areas and lakes. Because exclusion is economically feasible, a private supplier might actually come forward t o provide the good. For example, an enterprising individual might decide to
build a bridge and charge a crossing toll that would generate a stream o f revenue
more than adequate t o cover the cost o f construction-hence the label toll good.
Clearly, in these situations the problem is not supply per se. Rather, the problem is
twofold. First, the private supplier may not efficiently price the facility that is provided. Second, the private supplier may not provide the correct facility size to maximize social surplus [Qs in Figure 5.l(b)].
W i t h respect t o the pricing problem, first consider the case o f no congestion
(NEI), one in which we can be certain that the private supplier will not charge the
efficient price, which is zero. In the absence o f congestion, the social marginal cost o f
consumption is zero, so that any positive price inappropriately discourages use o f the
bridge. The shaded triangular area abc in panel (a) o f Figure 5.3 represents the deadweight loss that results from a positive price (or toll) PI with the demand schedule for
crossings given by D. From the diagram we can see that any positive price would involve some welfare loss. The reason is that i f a positive price is charged, some individuals who "should cross the bridge will be discouraged from doing so. W h y ? Because those individuals who would receive marginal benefits in excess of or equal
to, the marginal social cost o f consumption should cross. As the marginal social cost
o f consumption is zero (that is, lies along the horizontal axis), any individual who
would derive any positive benefit from the crossing should cross. Those who obtain
positive marginal benefits less than the price, however, would not choose t o cross.
The resulting deadweight loss may seem nebulous-the lost consumer surplus o f the
trips that will not take place because o f the t o l l b u t i t represents a misallocation o f
resources that, i f corrected, could lead to a Pareto improvement.
The analysis just presented is complicated i f the good displays congestion over
some range o f demand (NE2). Return to the bridge example. W e assumed that ca-

..
80

Rationales for Public Policy: Market Failures

Chap. 5

ply): the degree of rivalry in consumption; the extent o f excludability, or exclusiveness, in use; and the existence of congestion. The presence of nonrivalry, nonexcludability. or congestion arising from changes in levels of demand can lead to the failure
of markets to achieve Pareto efficiency. The presence of either nonrivalry or nonexcludability is a necessary condition for the existence of a public good market failure.

A Classification of Public Goods

Figure 5.2 presents the basic taxonomy of public goods with the rivalrous/
nonrivalrous distinction labeling the columns and excludability/nonexcludabilitydistinction labeling the rows. Additionally, the diagonals within the cells separate congested from uncongested cases. By the way, note that the definitions of public goods
that we provide differ starkly from a common usage of the term ''public good" to describe any good provided by government. In fact, governments, and indeed markets.
provide both public and private goods as defined.

RIVALROUS
NWI
/

Uncongested: Private Good

//
/

.//

NONRIVALROUS

NEl

/ //'
/

Uncongested: Toll G?od

NW2

/' NE2

No private supply at
/ / / /Congested:
efficient price of zero;/ /
///
Congested: underconsumption / /
Toll Good with
/
at any positive /,'
price.
1
Crowdirlg
//
Private Good wit17
/
Consltrnption E.rtenlnliQ'
/
/
/
Private supply can be
/
/
efficient
if price at marginal
Overconsumption because
1
social cost; peak-load pricing
consumers respond to price
required if congestion variable.
rather than marginal social cost / /
SE1
-I . . . / / /
/
Uncon_gested: "Free Good"
/
Uncongested: Pure Public G o o d / / /
/
1
SE2
S W 2 Private supply unlikely
//
Supply exceeds demand
//'
/ / Congested:
because exclusion not / / /
at zero price; no
d .
,/ /
Congested:
,/I
Open Access possible; some
a ~nefficiencyuntil
demand grows to / / '
and Coii1ii7on private supply in ,/'
U exceed supply / / / P r o p e p Reso~trces privileged and / ' ~ ~ i ~ b i ePublic
n t Good
with Coilsunlption
at zero price.,,'
Consumers respond to intermediate;'
%roUPS./ / /
Exfen7aIif)l
/ / marginal private cost rather
25.
than marginal social cost- !
///
//overconsumption. rent dissipation,
//
Overconsumption because
consumers ignore external cost.
.'/and underinvestment in presentation. / /
Market Supply

,//

./

.'/

.//

5
3
g

il

Figure 5.2 A Classificationof Goods: Private and Public

Public Goods

81

Rivalry, Excludability: Private Goods. The northwest (NW) cell defines private goods, characterized by both rivalry in consumption and excludability in
use: shoes, books, bread, and the other things we commonly purchase and own. In
the absence of congestion or other market failures, the self-interested actions of consumers and firms elicit and allocate these goods efficiently so that government intervention would have to be justified by some other rationale than the promotion of efficiency.
When the private good is congested (that is, it exhibits externalities o f consumption), market supply is generally not efficient. (In competitive markets the marginal social costs of the good equals not just the price-the marginal cost of production--but the sum of price and the marginal social costs of consumption.) Rather
than consider such situations as public goods market failures, it is more common to
treat them under the general heading o f externality market failures. Consequently,
we postpone consideration of category NW2 in Figure 5.2 to our d~scussionof external~
ties.
Nonrivalry, Excludability: Toll Goods. The northeast cell (NE) includes
those goods characterized by nonrivalry in consumption and excludability. Such
goods are often referred to as toll goods. Prominent examples include bridges and
roads that, once built, can carry significant levels o f traffic without crowding. Other
examples include goods that occur naturally such as wilderness areas and lakes. Because exclusion is economically feasjble, a private supplier might actually come forward to provide the good. For example, an enterprising individual might decide to
build a bridge and charge a crossing toll that would generate a stream of revenue
more than adequate to cover the cost of construction-hence the label toll good.
Clearly, in these situations the problem is not supply per se. Rather, the problem is
twofold. First, the private supplier may not efficiently price the facility that is provided. Second, the private supplier may not provide the correct facility size to maximize social surplus [Qxin Figure 5.l(b)J.
With respect to the pricing problem, first consider the case of no congestion
(NEI), one in which we can be certain that the private supplier will not charge the
efficient price, which is zero. In the absence ofcongestion, the social marginal cost o f
consumption is zero, so that any positive price inappropriately discourages use of the
bridge. The shaded triangular area abc in panel (a) of Figure 5.3 represents the deadweight loss that results from a positive price (or toll) PI with the demand schedule for
crossings given by D.From the diagram we can see that any positive price would involve some welfare loss. The reason is that i f a positive price is charged, some individuals who "should" cross the bridge will be discouraged from doing so. Why? Because those individuals who would receive marginal benefits in excess of: or equal
to, the marginal social cost of consumption should cross. As the marginal social cost
of consumption is zero (that is, lies along the horizontal axjs), any individual who
would derive any positive benefit from the crossing should cross. Those who obtain
positive marginal benefits less than the price, however, would not choose to cross.
The resulting deadweight loss may seem nebulous-the lost consumer surplus of the
trips that will not take place because o f the toll--but it represents a misallocation of
resources that, ifcorrected, could lead to a Pareto improvement.
The analysis just presented is complicated i f the good displays congestion over
some range of demand (NE2). Return to the bridge example. We assumed that ca-

Capacity
constraint

Marginal social cost of consumption: Ode

I .

(a) Social Surplus Loss with Positive Price in Absence of Congestion

Capacity
constraint

83

Public Goods

pacity exceeded demancCthis need not be the case. Consumption is typically uncongested up to some level o f demand, but as additional people use the good, the
marginal social costs o f consumption become positive. Goods such as bridges, roads,
and parks are potentially subject to congestion because o f physical capacity constraints. Whether or not they are actually congested depends on demand as well as
capacity. Panel (b) o f Figure 5.3 shows the marginal social costs o f consumption for
goods that are congested. A t low demand levels (for example, DL),the marginal cost
o f consumption is zero (in other words, consumption is uncongested), but at high
levels o f demand (for example, DH), consumption at a zero toll imposes marginal
costs on all users of the good. Line segments OJfg, andgh trace out the marginal social cost o f consumption for the good over the range o f possible consumption. Notice that these costs are imposed by the marginal consumer, not by the good itself. In
the case o f the bridge, for instance, the costs appear as the additional time it takes all
users to cross the bridge. I f additional users crowd onto the bridge, then quite conceivably marginal costs could become almost infinite: "Gridlock prevents anyone
from crossing the bridge until some users withdraw. The economically efficient price
is shown in panel (b) as PC.
Let us take a closer look at the incentives that lead to sociallv inefficient
crowding. Suppose that there are 999 automobiles on the bridge experiencing an average crossing time o f ten minutes. You are considering whether your auto should
be number 1,000 on the bridge. Unfortunately, given the capacity o f the bridge, your
trip will generate congestion-veryone
crossing the bridge, including yourself, will
be slowed down by one minute. In other words, the average crossing time o f the
users rises by one minute to eleven minutes. You will cross i f your marginal benefits
from crossing exceed your marginal costs (which in this case equal the new average
consumption cost o f eleven minutes for the group o f users). If you decide to cross,
however, the marginal social costs ofyour decision will be 1,010 minutes! (the eleven
minutes vou beardus the 999 additional minutes vour use inflicts on the other
,
;he social perspective, if everyoneZplacesthe same value on time
users). ~ h u sfrom
as you, then you should cross only i f the benefits that you receive from crossing exceed the cost o f 1.010 minutes o f delav.
In practice, nonrivalrous, excludable public goods that exhibit congestion involve quite complex pricing problems; however, the basic principle follows readily
from the above discussion. Let us assume that the congestion occurs at regular time
~eriodsas demand shifts over times o f the dav or seasons of the vear
(roads at rush
,
hour, for instance). Efficient allocation requires that the price charged users o f the
good equal the marginal costs imposed on other users during each period o f the day,
implying a zero price during uncongested periods (more generally, price should equal
the marginal cost o f production in the absence o f congestion) and some positive price
during congested periods (so called peak-load pricing).
Many toll goods are produced by private firms. The firms must pay the cost o f
producing the
with revenues from user fees. That is, the stre& o f revenues
from the fees must cover the cost o f construction and operation. T o generate the
necessary revenues, firms must set the tolls above the marginal social costs o f consumption. Indeed, as profit maximizers, they set tolls at levels that maximize their
rent [the area o f rectangle Pla QIO in Figure 5.3 (a)] given the demand schedule they
face. Thus, market failure results because the fees exclude users who would obtain
\

QuantityITime
Marginal social cost of consumption:Ofgh
(b) Appropriate Positive Price in Presence of Congestion

Figure 5.3 Toll Goods

Capacity
constraint

Public Goods

QuantityITime
Marginal social cost of consumption: Ode

(a) Social Surplus Loss with Positive Price in Absence of Congestion

Capacity
constraint

QuantitylTime
Marginal social cost of consumption: 0.fgh
(b) Appropriate Positive Price in Presence of Congestion
I

Figure 5.3 Toll Goods

83

pacity exceeded deman&this need not be the case. Consumption is typically uncongested up to some level o f demand, but as additional people use the good, the
marginal social costs o f consumption become positive. Goods such as bridges, roads.
and parks are potentially subject to congestion because o f physical capacity constraints. Whether or not they are actually congested depends on demand as well as
capacity. Panel (b) o f Figure 5.3 shows the marginal social costs o f consumption for
goods that are congested. At low demand levels (for example, DL),the marginal cost
o f consumption is zero (in other words, consumption is uncongested), but at high
levels o f demand (for example, DH), consumption at a zero toll imposes marginal
costs on all users o f the good. Line segments Offg, andgh trace out the marginal social cost o f consumption for the good over the range o f possible consumption. Notice that these costs are imposed by the marginal consumer, not by the good itself. In
the case o f the bridge, for instance, the costs appear as the additional time i t takes all
users to cross the bridge. If additional users crowd onto the bridge, then quite conceivably marginal costs could become almost infinite: "Gridlock prevents anyone
from crossing the bridge until some users withdraw. The economically efficient price
is shown in panel (b) as PC.
Let us take a closer look at the incentives that lead to socially inefficient
crowding. Suppose that there are 999 automobiles on the bridge experiencing an average crossing time o f ten minutes. You are considering whether your auto should
be number 1.000 on the bridge. Unfortunately, given the capacity o f the bridge, your
trip will generate congestion-veryone
crossing the bridge, including yourself, will
be slowed down by one minute. In other words, the average crossing time o f the
users rises by one minute to eleven minutes. You will cross if your marginal benefits
from crossing exceed your marginal costs (which in this case equal the new average
consumption cost o f eleven minutes for the group o f users). I f you decide to cross,
however, the marginal social costs o f our decision will be 1.010 minutes! (the eleven
minutes you bear plus the 999 additional minutes your use inflicts on the other
users). Thus, from the social perspective, if everyone places the same value on time
as you, then you should cross only i f the benefits that you receive from crossing exceed the cost of 1,010 minutes of delay.
In practice, nonrivalrous, excludable public goods that exhibit congestion involve quite complex pricing problems; however, the basic principle follows readily
from the above discussion. Let us assume that the congestion occurs at regular time
periods as demand shifts over times of the day or seasons o f the year (roads at rush
hour, for instance). Efficient allocation requires that the price charged users o f the
good equal the marginal costs imposed on other users during each period o f the day,
implying a zero price during uncongested periods (more generally, price should equal
the marginal cost o f production in the absence o f congestion) and some positive price
during congested periods (so called peak-load pricing).
Many toll goods are produced by private firms. The firms must pay the cost of
producing the goods w ~ t hrevenues from user fees. That is, the stream o f revenues
from the fees must cover the cost o f construction and operation. T o generate the
necessary revenues, firms must set the tolls above the marginal social costs of consumption. Indeed, as profit maximizers, they set tolls at levels that maximize their
rent [the area o f rectangle P,a Q,O in Figure 5.3 (a)] given the demand schedule they
face. Thus, market failure results because the fees exclude users who would obtain

Public Goods

.84

Rationales for Public Policy: Market Failures

Chap. 5

The privileged group case is illustrated in Figure 5.4, where three consumers
receive benefits from the good according t o their marginal benefit schedules. For example, the three might be owners of recreational facilities in a valley and the good
might be spraying t o control mosquitoes. The marginal benefit schedule o f person
three (D3)is high relative to the marginal benefit schedules o f persons one and two.
(By relatively high, w e mean that at any quantity, person three places a much higher
marginal value on having an additional unit than do the other t w o persons.) In fact, it
is sufficiently high that person three will be willing t o purchase Q3 o f the good no
matter what the other t w o persons do. (Person three's demand schedule intersects
with the supply schedule a t quantity Q3.)Of course, once person three purchases
Q3,neither person one nor person t w o will be willing to make additional purchases.
In effect, they free ride on person three's high demand. (We can speak o f person
three's marginal benefit schedule as a demand schedule because he will act as i f the
good were private, revealing his demand at various prices.) Despite the provision o f

higher marginal benefits than the marginal social costs that they impose. The magnitude o f the forgone net benefits determines the seriousness o f the market failure.
The problem o f inefficient scale arises in the case o f private supply because
firms seeking t o maximize profits anticipate charging tolls that restrict demand t o levels that can be accommodated b y smaller facilities. The result is that the facilities
built are too small from a social perspective. Further, the choice o f facility size determines the level o f demand at which congestion becomes relevant.

Nonrivalry, Nonexcludability: Pure and Ambient Public Goods.

W e n o w turn t o those goods that exhibit nonrivalrous consumption and where exclusion is not feasible-the
southeast (SE) quadrant o f Figure 5.2. When these
goods are uncongested, they are pure public goods. The classic examples o f such
public goods are defense and lighthouses. One o f the most important public goods in
modern societies is the generally available stock o f information that is valuable in production or consumption. W i t h certain exceptions, to be discussed below, pure public
goods will not be supplied at all by markets because o f the inability o f private
providers t o exclude those w h o do not pay for them. Contrast this w i t h the N E
quadrant, where there is likely t o be market provision, but a t a price that results in
deadweight losses-some positive price generates revenues t o cover cost&ut also
discourages some use involving zero marginal social costs o f consumption.
The number of persons w h o may potentially b e w f i t from a pure public good
can vary enormously, depending on the good: ranging from a particular streetlight
w i t h only a few individuals benefitting t o national defense w i t h all members o f the
polity presumably benefitting. Because benefits normally vary spatially or geographically (that is, benefits decline monotonically as one moves away from a particular
point on the map), w e commonly distinguish among local, regional, national, intemational, and even global public goods. While this is a convenient way o f grouping persons w h o receive benefits, i t is only one o f many potential ways. For example, persons who place positive values on wilderness areas in the Sierras may be spread all
over North America-indeed, all over the world. Some, or even most, o f those who
actually reside in the Sierras may not be included in this category because their private interests depend upon commercial or agricultural development o f the area
rather than upon preservation.
W e have already touched briefly on the major problem in the SE quadrant.
People who would actually receive some level o f positive benefits, if the good is provided, often do not have an incentive t o reveal honestly the magnitude o f these benefits: If contributions for the public good are t o be based on benefit levels, then individuals generally have an incentive t o understate their benefit levels; if contributions
are not tied t o benefit levels, then individuals may have an incentive to overstate
their benefit levels to obtain a larger supply. Typically, the larger the number o f beneficiaries, the less likely is any individual t o reveal his or her preferences. In such situations, private supply is unlikely. As Mancur Olson has pointed out, however, t w o
spectfic situations can result in market supply o f pure public goods.4 H e labels these
situations the privileged group and the intermediate group cases.

4Mancur Olson. The Logic of Collective Actlon (Cambridge, Mass.: Harvard University Press.

1973), pp. 43-52.

85

Socially optimal level of provision:

Q0
Private provision by person three:
Q3
Deadweight loss from underprovision: abc

Figure 5.4 Private Provision o f a Public Good: Privileged G r o u p

.84

Rationales for Public Policy: Market Failures

Chap. 5

higher marginal benefits than the marginal social costs that they impose. The magnitude o f the forgone net benefits determines the seriousness o f the market failure.
The problem o f inefficient scale arises in the case o f private supply because
firms seeking to maximize profits anticipate charging tolls that restrict demand to levels that can be accommodated by smaller facilities. The result is that the facilities
built are too small from a social perspective. Further, the choice o f facility size determines the level o f demand at which congestion becomes relevant.

Nonrivalry, Nonexcludability: Pure and Ambient Public Goods.

W e n o w turn t o those goods that exhibit nonrivalrous consumption and where exclusion is not feasible-the
southeast (SE) quadrant o f Figure 5.2. When these
goods are uncongested, they are pure public goods. The classic examples o f such
public goods are defense and lighthouses. One o f the most important public goods in
modern societies is the generally available stock o f information that is valuable in production or consumption. W i t h certain exceptions, t o be discussed below, pure public
goods will not be supplied at all by markets because o f the inability o f private
providers to exclude those who do not pay for them. Contrast this with the NE
quadrant, where there is likely to be market provision, but at a price that results in
deadweight losses-some positive price generates revenues t o cover cost$ut also
discourages some use involving zero marginal social costs o f consumption.
The number o f persons who may potentially bergfit from a pure public good
can vary enormously, depending on the good: ranging from a particular streetlight
with only a few individuals benefitting t o national defense with all members o f the
polity presumably benefitting. Because benefits normally vary spatially or geographically (that is, benefits decline monotonically as one moves away from a particular
point on the map), w e commonly distinguish among local, regional, national, international, and even global public goods. While this is a convenient way o f grouping persons who receive benefits, ~tis only one o f many potential ways. For example, persons w h o place positive values on wilderness areas in the Sierras may be spread all
over N o r t h America-indeed, all over the world. Some, or even most, o f those who
actually reside in the Sierras may not be included in this category because their private interests depend upon commercial or agricultural development o f the area
rather than upon preservation.
W e have already touched briefly on the major problem in the SE quadrant.
People who would actually receive some level o f positive benefits, i f the good is provided, often do not have an incentive t o reveal honestly the magnitude o f these benefits: I f contributions for the public good are to be based on benefit levels, then individuals generally have an incentive t o understate their benefit levels; i f contributions
are not tied to benefit levels, then individuals may have an incentive to overstate
their benefit levels to obtain a larger supply. Typically, the larger the number ofbeneficiaries, the less likely is any individual t o reveal his or her preferences. In such situations, private supply is unlikely. As Mancur Olson has pointed out, however, t w o
specific situations can result in market supply o f pure public goods.4 H e labels these
situations the privilegedgroup and the intermediate group cases.

4Mancur Olson. The log^ of Collective Actton (Cambridge. Mass.: Harvard University Press,
1973). pp. 43-52.

Public Goods

85

The privileged group case is illustrated in Figure 5.4, where three consumers
receive benefits from the good according t o their marginal benefit schedules. For example, the three might be owners o f recreational facilities in a valley and the good
might be spraying to control mosquitoes. The marginal benefit schedule o f person
three (03) is high relative t o the marginal benefit schedules of persons one and two.
(By relatively high, we mean that at any quantity, person three places a much higher
marginal value on having an additional unit than do the other t w o persons.) In fact, it
is sufficiently high that person three w ~ l be
l willing to purchase Q3 o f the good no
matter what the other t w o persons do. (Person three's demand schedule intersects
with the supply schedule at quantity Q3.) Of course, once person three purchases
Q3, neither person one nor person t w o will be willing to make additional purchases.
In effect, they free ride on person three's high demand. (We can speak o f person
three's marginal benefit schedule as a demand schedule because he will act as if the
good were private, revealing his demand at various prices.) Despite the provision o f

demand

Socially optimal level of provision:

Qo
Private provision by person three:
Q3
Deadweight loss from underprovision: abc

Figure 5.4 Private Provision of a Public Good: Privileged Group

Rationales for Public Policy: Market Failures

Chap. 5

Q, units of the public good. compared t o the socially efficient level Qo, social surplus
is lower by the area of triangle abc.
In this case, the demand of person three makes up such a large fraction of total
demand that the amount purchased (Q3) is fairly close t o the economically efficient
level ( Q o ) In this sense, the three persons form a privileged group. Even when no
d,
some
one person has a sufficiently high demand t o make a group ~ r i v i l e ~ ehowever,
provision of the good may result if the group is sufficiently small so that members can
negotiate directly among themselves. W e recognize such a group as "intermediate"
between privileged and unprivileged: t w o or more members may, for whatever reasons, voluntarily join together t o provide some of the good, although usually a t less
than the economically efficient level. lntemediate groups are generally small, or a t
least have a small number of members w h o account for a large fraction of total demand.
T h e situation just described closely resembles a positive externality (benefits
accruing t o third parties t o market transactions), which w e discuss in detail in the
next section. Clearly, if ~ e r s o n one
s and t w o did not agree t o finance jointly the public good a t the efficient level (Qo), they would nevertheless receive benefits from Q3.
the amount purchased by person three. Through the purchase of Q3, person three
receives the private consumer surplus benefit given by the area of triangle P3bd and
confers an external (to himself) consumer surplus benefit of area b c g on the other
group members. For analytic convenience, however, w e maintain a distinction between public goods and positive externalities. W e reStrict the definition of an externality t o those situations where provision of a good necessarily requires the joint
production of a private good and a public good. We reserve the;public good cclssification for those cases where there is no joint production. So, for example, mosquito
control poses a public good problem, while chemical waste produced as a by-product
of manufacturing poses an externality problem.
In many situations. large numbers of individuals would benefit from the provision of a public good where no small group receives a disproportionate share of total
benefits. Such "large-number" cases present the free rider problem with a
vengeance. In situations involving large numbers, each person's demand is extremely
small relative t o total demand and t o the cost of provision. T h e rational individual
compares his individual marginal benefits and costs. Taking as an example national
defense, the logic is likely t o be a s follows: My monetary contribution t o the financing of national defense will be infinites~mal;therefore, if I do not contribute and
everyone else does, the level of defense provided, from which I cannot be effectively
excluded, will be essentially the same as ~f1 did contribute. O n the other hand, if I
contribute and others d o not, national defense will not be provided anyway. Either
way, I am better off not contributing. (As w e will discuss below, free riding arises in
other contexts besides the market provision of public goods; it can also occur when
attempts are made t o supply the economically efficient quantity of the public good
through public sector mechanisms.)
T o summarize, then, the free rider problem exists because in the large numbers case it is usually impossible t o get persons t o reveal their true demand (marginal
benefit) schedules for the good (it is even difficult t o talk of individual demand schedules in this context because they are not generally observable). Even though all
would potentially benefit if all persons agreed t o contribute t o the financing of the
good so that their average contributions (in effect, the price they each paid per unit

Public Goods

87

supplied) just equaled their marginal benefits, self-interest in terms of personal costs
and benefits discourages honest participation.
T h e concept of free riding plays an important role in the theory of public
goods. Recently, however, there has been considerable debate among economists
about the practical significance of free riding. Both observational and experimental
evidence, however, suggest that free riding occurs, but perhaps not t o the level predicted by theory.5 John Ledyard, in a summary of the experimental literature, concludes: "(I) In one-shot trials and in the initial stages of finitely repeated trials, subjects generally provide contributions halfbay between the Pareto-efficient level and
the freeriding level, (2) Contributions decline with repetition, and (3) Face to face
communication improves the rate of c ~ n t r i b u t i o n . Another
"~
survey concludes that
higher individual marginal per capita returns from the public good raises contributions, but that free riding problems tend t o be worse in larger groups than in smaller
groups.7 In the context of relatively small groups that permit face-to-face contact,
such as neighborhood associations, social pressure may be brought t o bear t o make
failure t o contribute unpleasant.8 More generally, w e would expect a variety of voluntary organizations with less individual anonymity t o arise t o combat free riding.
There has also been considerable theoretical interest in pricing mechanisms that encourage people t o reveal their preferences truthfully, though they have as yet been
little used in practice.9
Thus far w e have not considered the issue of congestion in this SE quadrant.
Some goods are simply not congestible and can be placed t o the left of the diagonal
(SEI) within the SE quadrant. For instance, mosquito control, a local public good,
involves nonexcludability, nonrivalry in consumption, and noncongestibility; no matter how many people are added t o the area, the effectiveness of the eradication remains the same. Similarly, national defense, a national or international public good, in
general is not subject t o congestion. In contrast, nature lovers may experience disutility when they meet more than a few other hikers in a wilderness area.
We label nonrivalrous and nonexcludable goods that exhibit congestion, and
thus are appropriately placed into SE2, a s ambient public goods with consumption externality. Air and large bodies of water are prime examples of public goods that are
exogenously provided by nature. For all practical purposes, consumption (use) of

'see, for example, Linda Goetz. T . F. Glover, and 13. Blswas. "The Effects o f Group Size and Income on Contributions to the Corporation for Public Broadcasting." Publlc Chorce. Vol. 77, no. 2, 1993,
p p 407-14. For another revlew of the evidence, see also Robert C. Mitchell and Richard T. Carson,
Using Surveys to Value Public Gmds: The Conttngent Valuation Method (Washington, D.C.: Resources for
the Future. 1989), pp. 13349.
"ohn Ledyard. "Public G o d s : A Survey of Experimental Research." in John H . Kagel and Alv~n
E. Roth, eds., The Handbook of Experimental Economics (Princeton, N.J.: Princeton University Press,
1995). pp. 111-94, atp. 121.
7~ouglasD. Davis and Charles A Holt. Experimental Economrcs (Pr~nceton,N.J.: Princeton University Press. 1993). pp. 332-33.
'~homasS. McCaleb and Richard E. Wagner. "The Experimental Search for Free Riders: Some
Reflections and Observations." Public Choice, Vol. 47. no. 3, 1985, pp. 479-90.
an overview o f Vickrey, Clarke-Groves, and other demand revelation mechanisms, see Dennis C. Mueller, Publrc Cho~ce// (New York: Cambridge University Press, 1995), pp. 12434.

or

Rationales for Public Policy: Market Failures

Chap. 5

Q3 units o f the public good, compared t o the socially efficient level Qo,social surplus

is lower by the area o f triangle abc.

In this case, the demand o f person three makes up such a large fraction o f total
demand that the amount purchased (Q3) is fairly close t o the economically eficient
level (Qo). In this sense, the three persons form a privileged group. Even when no
one person has a sufficiently high demand t o make a group privileged, however, some
provision o f the good may result if the group is sufficiently small so that members can
negotiate directly among themselves. W e recognize such a group as "intermediate"
between privileged and unprivileged: t w o or more members may, for whatever reasons, voluntarily join together to provide some of the good, although usually at less
than the economically efficient level. Intermediate groups are generally small, or at
least have a small number o f members w h o account for a large fraction o f total demand.
The situation just described closely resembles a positive externality (benefits
accruing t o third parties t o market transactions), which w e discuss in detail in the
next section. Clearly, if persons one and t w o did not agree to finance jointly the public good at the eficient level (Q,), they would nevertheless receive benefits from Q3,
the amount purchased by person three. Through the purchase o f Q3, person three
receives the private consumer surplus benefit given by the area o f tri:ngle P3bd and
confers an external (to himself) consumer surplus benefit o f area bced on the other
group members. For analytic convenience, however, w e maintain a distinction between public goods and positive externalities. W e re'strict the definition o f an externality to those situations where provision o f a good necessarily requires the joint
production o f a private good and a public good. W e reserve the-public good classification for those cases where there is no joint production. So, for example, mosquito
control poses a public good problem, while chemical waste produced as a by-product
o f manufacturing poses an externality problem.
In many situations, large numbers o f individuals would benefit from the provision o f a public good where no small group receives a disproportionate share of total
benefits. Such "large-number" cases present the free rider problem w i t h a
vengeance. In situations involving large numbers, each person's demand is extremely
small relative t o total demand and t o the cost o f provision. T h e rational individual
compares his individual marginal benefits and costs. Taking as an example national
defense, the logic is likely t o be as follows: M y monetary contribution t o the financing o f national defense will be infinitesimal; therefore, if I do not contribute and
everyone else does, the level o f defense provided, from which 1 cannot be effectively
excluded, will be essentially the same as i f I did contribute. O n the other hand, if I
contribute and others do not, national defense will not be provided anyway. Either
way, I am better off not contributing. (As w e will discuss below, free riding arises in
other contexts besides the market provision o f public goods; it can also occur when
attempts are made t o supply the economically efficient quantity o f the public good
through public sector mechanisms.)
T o summarize, then, the free rider problem exists because in the large numbers case it is usually impossible t o get persons t o reveal their true demand (marginal
benefit) schedules for the good (it is even difficult t o talk o f individual demand schedules in this context because they are not generally observable). Even though all
would potentially benefit i f all persons agreed t o contribute t o the financing o f the
good so that their average contributions (in effect, the price they each paid per unit

Public Goods
supplied) just equaled their marginal benefits, self-interest in terms o f personal costs

and benefits discourages honest participation.

The concept of free riding plays an important role in the theory o f public
goods. Recently, however, there has been considerable debate among economists
about the practical significance o f free riding. Both observational and experimental
evidence, however, suggest that free riding occurs, but perhaps not to the level pre. ~ Ledyard, in a summary o f the experimental literature, condicted by t h e ~ r yJohn
cludes: "(1) In one-shot trials and in the initial stages o f finitely repeated trials, subjects generally provide contributions h a l h a y between the Pareto-efficient level and
the freeriding level, (2) Contributions decline with repetition, and (3) Face t o face
communication improves the rate of c o n t r i b ~ t i o n . Another
"~
survey concludes that
higher individual marginal per capita returns from the public good raises contributions, but that free riding problems tend t o be worse inlarger groups than in smaller
groups.7 In the context of relatively small groups that permit face-to-face contact,
such as neighborhood associations, social pressure may be brought t o bear t o make
failure t o contribute unpleasant.8 More generally, w e would expect a variety o f voluntary organizations with less individual anonymity t o arise t o combat free riding.
There has also been considerable theoretical interest in pricing mechanisms that encourage people t o reveal their preferences truthhlly, though they have as yet been
little used in practice.9
Thus far w e have not considered the issue o f congestion in this SE quadrant.
Some goods are simply not congestible and can be placed t o the left o f the diagonal
(SEI) within the SE quadrant. For instance, mosquito control, a local public good,
involves nonexcludability, nonrivalry in consumption, and noncongestibility; no matter how many people are added t o the area, the effectiveness o f the eradication remains the same. Similarly, national defense, a national or international public good, in
general IS not subject t o congestion. In contrast, nature lovers may experience disutility when they meet more than a few other hikers in a wilderness area.
W e label nonrivalrous and nonexcludable goods that exhibit congestion, and
thus are appropriately placed into SE2, as ambient public goods with consumption externality. Air and large bodies o f water are prime examples o f public goods that are
exogenously provided by nature. For all practical purposes, consumption (use) o f

5See,for example, Linda Goetz. T. F. Glover, and B. Biswas, "The Effects of Group Size and Income on Contributions to the corporation for Public Broadcasting." Public Choice. Vol. 77. no. 2, 1993.
pp 407-14. For another review o f the evidence, see also Robert C. Mitchell and Richard T. Carson,
Using Surveys to Value Public Goods: The Contingent Valuat~onMethod (Washington, D.C.: Resources for
the Future. 1989). pp. 13349.
6 ~ o h Ledyard,
n
"Public Gods: A Survey of Experimental Research." In John H . Kagel and Alvin
E. Roth, eds.. The Handbook ofExperimenta1 Economics (Princeton, N.J.: Princeton University Press,
1995). pp. 1 1 1-94, at p. 121.
7~ouglasD. Davis and Charles A. Holt. Experimental Econom;cs (Princeton. N.J.: Princeton University Press. 1993), pp. 332-33.
'Thomas S.McCaleb and Richard E. Wagner. "The Expenmental Search for Free Riders: Some
Reflections and Observations." Public Choice, Vol. 47, no. 3, 1985, pp. 479-90.
'For an overview o f Vickrey, Clarke-Groves, and other demand revelation mechanisms, see Dennis C. Mueller, Public Choice // (New York: Cambridge University Press. 1995), pp. 12434.

Rationales for Public Policy: Market Failures

Chap. 5

these goods is nonrivalrous-more than one person can use the same unit for such
purposes as disposing of pollutants.'0 In other words, consumption of the resource
(via pollution) typically imposes no Pareto-relevant impact until some threshold, or
ambient carrying capacity, has been crossed. Exclusion in many market contexts is
impossible, or a t least extremely costly, because access t o the resources is possible
from many different locations. For example, pollutants can be discharged into an air
shed from any location under it (or even upwind of it!).
Relatively few goods fall into category SE2. T h e reason is that as congestion
sets in, it often becomes economically feasible t o exclude users so that goods that
might otherwise be in SE2 are better placed in NE2 instead. For example, returning
to wilderness hiking, once the density of hikers becomes very large, it may become
economically feasible for a private owner t o issue passes that can be enforced by
spot checks. Wilderness in this context is an ambient public good only over the range
of use between the onset of crowding and the reaching of a density where the pass
system becomes economically feasible. T h e efficiency problems associated with
many of the goods that do fall into category SE2 can alternatively be viewed as market failures due to externalities. So, for example, the carrying capacity of a body of
water can be viewed as an ambient public good that suffers from overconsumption.
Alternatively, the pollution that the body of water receives can be viewed as an externality of some other good that is overproduced.

Rivalry, Nonexcludability: Open ~ c c e s ; , Common Property Resources, and Free Goods. Consider goods in the southwest quadrant, where
consumption is rivalrous, but where exclusion is not economically feasible-in other
words, there is open access t o the good. W e should stress that in this quadrant w e
are dealing with goods that are rivalrous in consumption. Trees, fish, buffalo, oil, and
pasture land are all rivalrous in consumption: If I take the hide from a buffalo, for instance, that particular hide is no longer available for you t o take. Specifically, open
access means that anyone who can seize units of the good controls their use. W e
normally think of open access in terms of unrestricted entry of new users. However,
open access also involves unrestr~cteduse by a fixed number of individuals who already have access. But, in these circumstances it is plausible t o talk in terms of ownership--a fixed number of users may share collective property rights t o the good.
Given this property right, they have an incentive, and possibly the means, t o reduce
or eliminate overconsumption.
N o immediate market failure appears in those cases in which the good is naturally occurring and where supply exceeds demand at zero price. A s anyone can take
these goods without interfering with anyone else's use, w e refer t o them as free
goods (SWI). Thus, although they are theoretically rivalrous in consumption, from
an efficiency perspective they are not because of the excess supply.

Public G o o d s

89

In situations in which demand is higher so that there is no excess supply a t

zero price, w e have what is usually referred t o as an open access resource (SW2).
When access is limited t o a defined group of potential users, in other words, entry is
restricted, and the users own the good in common, there is common property owners h k . T h e distinction between open access and common property situations requires
some elaboration. In the case of common property, the limiting of access to a defined
set of persons opens the possibility of self-governance among them that reduces or
eliminates inefficiencies." In the case of open access, however, the threat of new entrants effectively eliminates the possibility of self-governance. Even in cases of common property, individually rational behavior by members of the defined group can
lead t o inefficiency in a way that may end up being indistinguishable from open access-in such cases common property results in a common property resource problem.
Consequently, although much of the following discussion is framed in terms of open
access, it is generally relevant t o common property as well.
Market failure arises in the open access case from the "infeasibility" of exclusion. Why the quotation marks? Because, as w e will see, open access problems
often occur in situations in which institutional features rather than the inherent nature of the goods make exclusion infeasible. For example, in most countries, oil does
not suffer from open access because their governments have adopted laws that keep
and enforce exclusive property rights t o subsurface resources for the governments
themselves. In the United States, however, oil has often suffered from the open access problem because of the "rule of capture," a legal doctrine that gives most subsurface rights t o the owner of the surface property. When different people own separate pieces of property over the same pool of oil, the rule of capture prevents
exclusion. Unless all the owners agree t o treat their combined property as a unit, the
common reservoir of oil will be extracted too quickly.
Nonexcludability leads t o economically inefficient overconsumption of rivalrous goods. It can also lead t o underinvestment in preserving the stock of goods or t o
overinvestment in capital used t o capture the good.'' Naturally occurring resources
are especially susceptible t o the open access problem. Persons with access t o the resource realize that what they do not consume will be consumed by someone else.
Each person, therefore, has an incentive t o consume the resource a t a faster rate
than if he or she had exclusive ownership. For instance, deforestation often results
when a population relies on a forest as a source of firewood. Further, the availability
of underpriced firewood does not give users the appropriate incentive t o Invest in
stoves that use less wood, and the fact that anyone can cut and gather wood discourages individuals from replanting o r nurturing trees.
Figure 5.5 illustrates the efficiency losses associated with overconsumption
when there is open access t o a resource. T h e marginal social benefit schedule (MSB)
represents the horizontal summation (remember, w e are dealing w ~ t ha rivalrous
good) of all the demand schedules of individuals. The economically efficient level of

"Other sorts o f consumpt~onmay be rivalrous. For example, taking water from a river for irrigation is rivalrous: Assuming congestion (positive marginal social costs o f consumption), the water should be
treated as an open access resource (SW2 in F~gure5.2) rather than as an amb~entpublic g o d . The same
river, however. might be nonrivalrous with respect to its capacity to carry away wastes. See Robert H

"Gary D.Libecap, Controct~ngforProperty Rrghts (New York: Cambridge University Press, !989)
and Elinor Ostrom, Governing the Commons: The Evolutron of Institutionsfor Collectrve Actron (New York:
t~
1990).
Cambridge U n i v e r s ~Press,

Haveman, "Common Property, Congest~on,and Environmental Pollution," Quarterly Journal of Economics. Vol. 87, no. 2. 1973, pp. 278-87.

"See Michael B Wallace. "Managing Resources That Are Common Property: From Kathmandu
to Capitol HIII," Journal of Poiicy Anolysrs and Management, Vol. 2, no. 2, 1983, pp 220-37.

Rationales for Public Policy: Market Failures

Chap. 5

these goods is nonrivalrous-more than one person can use the same unit for such
purposes as disposing o f pollutants.10In other words, consumption o f the resource
(via pollution) typically imposes no Pareto-relevant impact until some threshold, or
ambient carrying capacity, has been crossed. Exclusion in many market contexts is
impossible, or at least extremely costly, because access to the resources is possible
from many different locations. For example, pollutants can be discharged into an air
shed from any location under i t (or even upwind o f it!).
Relatively few goods fall into category SE2. The reason is that as congestion
sets in, i t often becomes economically feasible to exclude users so that goods that
might otherwise be in SE2 are better placed in N E 2 instead. For example, returning
t o wilderness hiking, once the density o f hikers becomes very large, it may become
economically feasible for a private owner to issue passes that can be enforced by
spot checks. Wilderness in this context is an ambient public good only over the range
o f use between the onset o f crowding and the reaching o f a density where the pass
system becomes economically feasible. The efficiency problems associated with
many o f the goods that do fall into category SE2 can alternatively be viewed as market failures due t o externalities. So, for example, the carrying capacity o f a body o f
water can be viewed as an ambient public good that suffers from overconsumption.
Alternatively, the pollution that the body o f water receives can be viewed as an externality o f some other good that is overproduced.

Rivalry, Nonexcludability: Open ~cces;, Common Property Resources, and Free Goods. Consider goods in the southwest quadrant, where
consumption is rivalrous, but where exclusion is not economically f e a s i b l e i n other
words, there is open access t o the good. W e should stress that in this quadrant we
are deal~ngwith goods that are rivalrous in consumption. Trees, fish, buffalo, oil, and
pasture land are all rivalrous in consumption: If I take the hide from a buffalo, for instance, that particular hide is no longer available for you t o take. Specifically, open
access means that anyone who can seize units o f the good controls their use. W e
normally think o f open access in terms o f unrestricted entry o f new users. However,
open access also involves unrestricted use by a fixed number o f individuals who already have access. But, in these circumstances it is plausible t o talk in terms o f ownership-a fixed number o f users may share collective property rights to the good.
Given this property right, they have an incentive, and possibly the means, to reduce
o r eliminate overconsumption.
N o immediate market failure appears in those cases in which the good is naturally occurring and where supply exceeds demand at zero price. As anyone can take
these goods without interfering with anyone else's use, we refer t o them as free
goods (SWI). Thus, although they are theoretically rivalrous in consumption, from
an efficiency perspect~vethey are not because o f the excess supply.

"Other sorts of consumption may be rivalrous. For example, taking water from a river for irrigation is rivalrous: Assurn~ngcongestion (positive marginal social costs o f consumption), the water should be
treated as an open access resource (SW2 in Figure 5.2) rather than as an ambient public good. The same
respect to its capacity to carry away wastes. See Robert H.
river, however, might be nonrivalrous w ~ t h
Haveman, "Common Property, Congestion. and Environmental Pollution," Quarterly Journal of Economics. Vol. 87, no. 2.1973. pp. 278-87.

Public Goods

89

In situations in which demand is higher so that there is no excess supply at

zero price, w e have what is usually referred to as an open access resource (SW2).
When access is limited t o a defined group o f potential users, in other words, entry is
restricted, and the users own the good in common, there is common property ownership. The distinction between open access and common property situations requires
some elaboration. In the case o f common property, the limiting o f access t o a defined
set o f persons opens the possibility o f self-governance among them that reduces or
eliminates inefficiencies." In the case o f open access, however, the threat o f new entrants effectively eliminates the possibility o f silf-governance.
Even in cases o f common property, individually rational behavior by members of the defined group can
lead to inefficiency in a way that may end up being indistinguishable from open access-in such cases common property results in a common property resource problem.
Consequently, although much o f the following discussion is framed in terms o f open
access, it is generally relevant to common property as well.
Market failure arises in the open access case from the "infeasibility" o f exclusion. W h y the quotation marks? Because, as w e will see, open access problems
often occur in situations in which institutional features rather than the inherent nature o f the goods make exclusion infeasible. For example, in most countries, oil does
not suffer from open access because their governments have adopted laws that keep
and enforce exclusive property rights t o subsurface resources for the governments
themselves. In the United States, however, oil has often suffered from the open access problem because o f the "rule o f capture," a legal doctrine that gives most subsurface rights t o the owner o f the surface property. When different people own separate pieces o f property over the same pool o f oil, the rule o f capture prevents
exclusion. Unless all the owners agree t o treat their combined property as a unit, the
common reservoir o f oil will be extracted too quickly.
Nonexcludability leads to economically inefficient overconsumption o f rivalrous goods. It can also lead t o underinvestment in preserving the stock o f goods or to
overinvestment in capital used t o capture the good.'2 Naturally occurring resources
are especially suscept~blet o the open access problem. Persons with access to the resource realize that what they do not consume will be consumed by someone else.
Each person, therefore, has an incentive t o consume the resource at a faster rate
than if he or she had exclusive ownership. For instance, deforestation often results
when a population relies on a forest as a source offirewood. Further, the availability
of underpriced firewood does not give users the appropriate incentive t o invest in
stoves that use less wood, and the fact that anyone can cut and gather wood discourages individuals from replanting or nurturing trees.
Figure 5.5 illustrates the efficiency losses associated with overconsumption
when there is open access t o a resource. The marginal social benefit schedule (MSB)
represents the horizontal summation (remember, we are dealing with a rivalrous
good) o f all the demand schedules o f individuals. The economically efficient level o f

"Gary D. Libecap, Contracting for Property Rights (New York: Cambridge University Press. !989)
and Elinor Ostrom, Governing the Commons: The Evolution of lnstitut~onsfor Collective Actron (New York.
Cambridge University Press. 1990).
"See Michael B. Wallace. "Managing Resources That Are Common Property: From Kathmandu
to Capitol Hill," Journalof Pol~cyAnalys~sandManogement. Vol. 2 , no. 2. 1983, pp. 220-37.

Chap. 5

Rationales for Public Policy: Market Failures

MSC

MPC ' = ASC

I
I

'MSB

(demand)

M P C = Marginal private cost

A4SC = Marginal social cost
A S C = Average social cost
Qo (MSB = M S C )
Socially optimal level of consumption:
Open access resource level of consumption: QoA
Deadweight loss from overconsumption:
nbc

Figure 5.5

Overconsumption of O p e n Access Resources

Qo, results when marginal social cost (MSC) equals marginal social
benefit (MSC = MSB). Each individual, however, takes account o f only the costs
that he or she directly bears, that is, marginal private costs (MPC). This private marginal cost turns out t o be the avsrage cost for all demanders (ASC) i f marginal social
cost is borne equally (that is, averaged) among consumers. W i t h everyone rationally
treating average group cost as their marginal cost, equilibrium consumption will be at
QOA, which is greater than the economically efficient level Qo.The shaded triangle
abc measures the loss in social surplus that results from the overconsumption.
A simple example may help clarify why ~ndividualsin open access and common
property situations have an incentive to respond to marginal private cost rather than
marginal social cost. Imagine that you are in a restaurant with a group o f ten people
who have agreed t o split the bill evenly. If you were paying your own tab, then you
would not order the fancy dessert costing $10 unless you expected to get at least
$10 worth of value from eating it. But because the actual cost t o you o f the dessert

91

Public Goods

will be $1 (the average increase in your bill, and for the bills o f everyone else in the
g r o u p $ l O divided by ten people), you would be rational (ignoring calories, your remaining stomach capacity, and social pressure) to order it as long as i t would give
you at least one more dollar in value. You m~ghtcontinue ordering desserts until the
value you placed on one more fell t o $I, your marginal private cost and the group average cost. In other words, your out-of-pocket cost for an additional dessert is the
increment t o your bill. It IS the cost o f the dessert split equally among the members
o f the group. But this result is clearly inefficient because you and everyone else in the
group could be made better o f f i f you refrained from ordering the last dessert in return for a payment from the others o f any amount between $I (your marginal private
cost) and $9 (the difference between the marginal social cost and your marginal private cost). Remember that the problem arises here because you have access to
items on the menu at below their real social cost.
Note t w o things about this example. First, as the number o f people in the
group is restricted to ten, this is a common property situation in that membership in
the group is closed. However, remember that we defined open access as including
"unrestricted use by those who already have access." If anyone could freely join the
group and join in the tab splitting, it would be a pure open access situation. As group
size increases, the divergence between marginal private and marginal group (social)
costs increases, so that over-ordering increases. Second, in this example, overordering is mitigated by the fact that participants must consume the food at the
table. If the participants could take doggie bags, or even resell ordered food, then the
incentives to over-order would be even greater. Natural resource extractors, who
typically sell seized units rather than consume them, are thus like those who take
doggie bags.
Modeling common property as a game between t w o potential users makes
clear how individual incentives lead to overexploitation from a social perspective.
Imagine that two ranchers have access to a grassland. Each must decide how many
head o f cattle to graze on the grassland not knowing how many head the other will
graze. For simplicity, imagine that the choices are either 50 or 100 head o f cattle.
These choices are the "strategies" available t o the ranchers in this game. Each pair
o f strategies, shown as the cells in Figure 5.6, yield pairs o f payoffs t o the t w o ranchers given as (Payoff to Rancher I, Payoff to Rancher 2). If each chooses to graze 50
head, then each earns a profit o f $1,000 because the total herd size o f 100 can be ac-

consumption,
1
1

Rancher 2
50 Head

100 Head

50 Head
Rancher 1
100 Head

Figure 5.6 Choice o f Herd Size as a Prisoner's Dilemma

90

Rationales for Public Policy: Market Failures

Chap. 5

MSC

MPC = ASC

I
I

I
I

'
MSB
(demand)

MPC = Marginal private cost

MSC = Marginal social cost
ASC = Average social cost
Socially optimal level of consumption:
Qo (MSB = MSC)
Open access resource level of consumption: QoA
Deadweight loss from overconsumption:
obc

Figure 5.5 Overconsumption of Open Access Resources

consumption, Qo, results when marginal social cost (MSC) equals marginal social
benefit (MSC = MSB). Each individual, however, takes account of only the costs
that he or she directly bears, that is, marginal private costs (MPC).This private marginal cost turns out to be the average cost for all demanders (ASC) if marginal social
cost is borne equally (that is, averaged) among consumers. With everyone rationally
treating average group cost as their marginal cost, equilibrium consumption will be at
QOA, which is greater than the economically efficient level Qo.The shaded triangle
obc measures the loss in social surplus that results from the overconsumption.
A simple example may help clarify why individuals in open access and common
property situations have an incentive to respond to marginal private cost rather than
marginal social cost. lmagine that you are in a restaurant mriih a group of ten people
who have agreed to split the bill evenly. If you were paying your own tab, then you
would not order the fancy dessert costing $10 unless you expected to get at least
$10 worth of value from eating it. But because the actual cost to you of the dessert

91

Public Goods

will be $I (the average increase in your bill, and for the bills of everyone else in the
group$lO divided by ten people), you would be rational (ignoring calories, your remaining stomach capacity, and social pressure) to order it as long as it would give
you at least one more dollar in value. You might continue ordering desserts until the
value you placed on one more fell to $I, your marginal private cost and the group average cost. In other words, your out-of-pocket cost for an additional dessert is the
increment to your bill. It is the cost of the dessert split equally among the members
of the group. But this result is clearly inefficient because you and everyone else in the
group could be made better off if you refrained from ordering the last dessert in return for a payment from the others of any amount between $I (your marginal private
cost) and $9 (the difference between the marginal social cost and your marginal private cost). Remember that the problem arises here because you have access to
items on the menu at below their real social cost.
Note two things about this example. First, as the number of people in the
group is restricted to ten, this is a common property situation in that membership in
the group is closed. However, remember that we defined open access as including
"unrestricted use by those who already have access." If anyone could freely join the
group and join in the tab splitting, it would be a pure open access situation. As group
size increases, the divergence between marginal private and marginal group (social)
costs increases, so that over-ordering increases. Second, in this example, overordering is mitigated by the fact that participants must consume the food at the
table. If the participants could take doggie bags, or even resell ordered food, then the
incentives to over-order would be even greater. Natural resource extractors, who
typically sell seized units rather than consume them, are thus like those who take
doggie bags.
Modeling common property as a game between two potential users makes
clear how individual incentives lead to overexploitation from a social perspective.
lmagine that two ranchers have access to a grassland. Each must decide how many
head of cattle to graze on the grassland not knowing how many head the other will
graze. For simplicity, imagine that the choices are either 50 or 100 head of cattle.
These choices are the "strategies" available to the ranchers in this game. Each pair
of strategies, shown as the cells in Figure 5.6, yield pairs of payoffs to the two ranchers given as (Payoff to Rancher I , Payoff to Rancher 2). If each chooses to graze 50
head, then each earns a profit of $1,000because the total herd size of I00 can be ac-

Rancher 2
50 Head

100 Head

50 Head
Rancher 1
100 Head

Figure 5.6 Choice of Herd Size as a Prisoner's Dilemma

Rationales for Public Policy: Market Failures

Chap. 5

commodated by the field. If each chooses to graze 100 head, then each earns a profit
o f only $700 because the total herd size is much too large for the field so that the
cattle gain too little weight. If Rancher I grazes 100 head and Rancher 2 grazes 50
head, then Rancher 1 earns a profit o f $1,200 and Rancher 2 a profit of only $600. If
Rancher I grazes 50 and Rancher 2 grazes 100, then i t is Rancher 2 w h o earns the
$1,200 and Rancher I the $600. In these strategy combinations, the herd size is too
big for the field, b u t the lost weight per head does not offset the advantage t o the
more aggressive rancher o f grazing the extra 50 head.
A n important basis for predicting behavior in a game is the Nash equilibrium. In
a two-person game, a pair o f strategies is a Nash equilibrium if, given the strategy o f
the other player, neither player wishes t o change strategies. In the game at hand, i t is
clear that each rancher restricting his herd size to 50 head is not a Nash equilibrium-ither
rancher could raise his profit from $1,000 to $1,200 b y switching t o 100
head. But i f only one switched t o 100 head, the other could raise his profits from
$600 to $700. Only when both choose 100 head would neither player have an incentive t o move back to 50 head. Thus, the only Nash equilibrium in this game is for
both t o choose herds o f 100 head. That this equilibrium is Pareto inefficient is
clear-ach
would be made better-off i f he chose a herd o f 50 head.
Games w i t h similar structure, called Prisoner's Dilemmas, are widely used by
. ' ~ are called noncooperative
social scientists t o model problems o f c ~ o ~ e r a t i o nThey
games because o f the assumption that the players cannot make b~ndingcommitments t o strategies before they must be chosen. If i t were possible t o make binding
commitments, then w e could imagine the ranchers cooperating b y agreeing to limit
their heard sizes t o 50 head before strategies are chosen. As this gape is only played
one time, it is called a single-play game. One could imag~nethat the ranchers faced
the problem o f choosing heard sizes each year. I t might then be reasonable to model
their interaction as a repeatcdgame consisting o f successive plays o f the single-play,
or stage, game. If the repeated game consists o f an infinite, or uncertain, number o f
repetitions, and players give sufficient weight t o future payoffs relative to current
payoffs, then cooperative equilibria may emerge that involve each repeatedly choosing strateg~esthat would not be equilibria in the stage game. (We return t o these
ideas when w e discuss corporate culture and leadership in Chapter 14.)
Natural resources (both renewable and nonrenewable) have the potential for
yielding scarcity rents-returns in excess o f the cost of production-because In instances o f open access, these rents may be completely dissipated. In Figure 5.5, for
instance, consumption at the economically efficient level Qowould yield rent equal
t o the area o f rectangle PoadACo.The economically efficient harvesting o f salmon,
for example, requires catch limits that keep the market price above the marginal
costs o f harvesting. In the absence o f exclusion, however, these rents may well be
completely dissipated.I4 ( A t consumption level QoA In Figure 5.5, there is no rent.)
The reason is that fishermen will continue t o enter the industry (and those already in

"For introductions to game theory, see James D. Morrow, Game Theoryfor Politico1 Scientists
(Princeton. N . J.: Princeton University Press. 1994) and Martin J. Osborne and Ariel Rubenstein, A
Course in Game Theory (Cambridge, Mass.: M I T Press, 1994).
I4See, for example, L. G. Anderson. The Econom~csof Flshery Management (Baltimore: Johns
Hopkins University Press. 1977).

Public Goods

93

it will fish more intensively) as long as marginal private benefits-the rents they can
capture-xceed
marginal private costs. Just as in the restaurant example, each fisherman will ignore the marginal costs that his behavior imposes on other fishermen. If
every fisherman is equally efficient, then his marginal private cost equals the average
cost for the fishermen as a group. The question o f h o w rent should be distributed is
usually one o f the most contentious Issues in public policy. For example, h o w should
the catch limits for salmon be divided among commercial, sport, and native fishermen? Nevertheless, from the perspective o f economic efficiency, someone should
receive the scarcity rent, rather than allowing it to be wasted.
Note that demand for a resource may be sufficiently low so that i t remains uncongested at zero price. Increases in demand, however, may push the resource from
a free good (SWI) t o an open access good (SW2). Historically, free goods have been
fairly common in North America, includ~ngsuch resources as buffalo, forests,
aquifer water, and rangeland. Typically, however, the free goods eventually disappeared as demand increased and excess supply was eliminated.I5 Open access often
led t o rapid depletion and, in some cases, near destruction o f the resource before effective exclusion was achieved. For example, open access permitted destruction o f
the Michigan, Wisconsin, and Minnesota pine forest^.'^ Nonexcludibility continues
to be at the heart of many water use problems in the western United States.17When
animal populations are treated as open access resources, the final result o f overconsumption may be species extinction; such a result has already occurred with certain
birds, and other animals w i t h valuable skin or fur.
Thus far w e have not specified the meaning o f "feasible" exclusion. Many o f
the goods we have given as examples appear not t o be inherently nonexcludible. Indeed, one of the most famous historical examples o f an open access resourcesheep and cattle grazing on the medieval English pasture-was "solved," willy-nilly,
without overt government intervention, by the enclosure movement, which secured
property rights for estate holders. Similar enclosures appear to be currently taking
place on tribal, historically open-access, lands in parts o f Africa.
I t is useful to dichotomize open access and common property problems into
those that are structural (where aspects o f the goods preclude economically feasible
exclusion mechanisms) and those that are institutional (where economically efficient
exclusion mechanisms are feasible but the distribution o f property rights precludes
the~rimplementation). The averaging o f restaurant bills, previously discussed, serves
as an excellent illustration of an institutional common property problem. W e can
imagine an obvious exclusion mechanism that w e know from experience is economi-

"ln the case of the American buffalo, the opening of the rallroad fac~l~tated
the huntlng and transportation of hldes at much lower costs so that what had previously been a free good became an open access resource. For an economic analysis, see John Hanner, "Government Response to the Buffalo Hide
Trade, 1871-1883," Journalof Law and Economrcs, Vol. 24, no. 2, 1981, pp 239-71.
'"ndrew Dana and John Baden. "The New Resource Econom~cs:Toward an ldeolog~calSynthesis," Policy Studies Journal, Vol. 14, no. 2, 1985, pp. 2 3 3 4 3 .
I7B. Delworth Gardner, "Institutional lmped~mentsto Efficient Water Allocation," Pol;cy Studies
Revrew, Vol. 5, no. 2, 1985, p p 353-63. See also W~lliarnBlomquist and Elinor Ostrom, "Institutional
Capacity and the Resolution of a Commons D~lemma,"Polrcy Studies Revtew. Vol. 5 , no. 2, 1985, pp,
283-93.

92'

Rationales for Public Policy: Market Failures

Chap. 5

comrnodated by the field. I f each chooses to graze 100 head, then each earns a profit
o f only $700 because the total herd size is much too large for the field so that the
cattle gain too little weight. If Rancher I grazes I00 head and Rancher 2 grazes 50
head, then Rancher 1 earns a profit o f $1,200 and Rancher 2 a profit o f only $600. If
Rancher I grazes 50 and Rancher 2 grazes 100, then it is Rancher 2 who earns the
$1,200 and Rancher 1 the $600. In these strategy combinations, the herd size is too
big for the field, but the lost weight per head does not offset the advantage to the
more aggressive rancher o f grazing the extra 50 head.
A n important basis for predicting behavior in a game is the Nash equilibrium. In
a two-person game, a pair o f strategies is a Nash equilibrium if: given the strategy o f
the other player, neither player wishes t o change strategies. In the game at hand, it is
clear that each rancher restricting his herd size to 50 head is not a Nash equilibrium-ither
rancher could raise his profit from $1,000 t o $1,200 by switching t o 100
head. But if only one switched to 100 head, the other could raise his profits from
$600 t o $700. Only when both choose 100 head would neither player have an incentive t o move back to 50 head. Thus, the only Nash equilibrium in this game is for
both t o choose herds o f 100 head. That this equilibr~umis Pareto inefficient is
clear-ach
would be made better-off i f he chose a herd o f 50 head.
Games with similar structure, called Prisoner's Dilemmas, are widelv used by
social sc~entiststo model problems o f cooperation.13They are called noncooperative
games because o f the assumption that the players cannot make binding commitments t o strategies before they must be chosen. If i t w6re possible t o make binding
commitments, then w e could imagine the ranchers cooperating by agreeing to limit
their heard sizes to 50 head before strategies are chosen. As this gape is only played
one time, i t is called a single-play game. One could imagine that the ranchers faced
the problem o f choosing heard sizes each year. I t might then be reasonable t o model
their interaction as a repeated game consisting o f successive plays o f the single-play,
or stage, game. If the repeated game consists o f an infinite, or uncertain, number o f
repetitions, and players give sufficient weight to future payoffs relative t o current
payoffs, then cooperative equilibria may emerge that involve each repeatedly choosing strategies that would not be equil~briain the stage game. (We return to these
ideas when w e discuss corporate culture and leadership in Chapter 14.)
Natural resources (both renewable and nonrenewable) have the potential for
yielding scarcity rents-returns in excess o f the cost o f production--because in instances o f open access, these rents may be completely dissipated. In Figure 5.5, for
instance, consumption at the economically efficient level Qo would yield rent equal
to the area o f rectangle PoadACo. The economically efficient harvesting o f salmon,
for example, requires catch limits that keep the market price above the marginal
costs o f harvesting. In the absence o f exclusion, however, these rents may well be
completely d ~ s s ~ ~ a t e( A
d .t ' consumption
~
level QOA in Figure 5.5, there is no rent.)
The reason is that fishermen will continue to enter the industry (and those already in

I 3 ~ ointroductions
r
to game theory, see James D. Morrow, Game Theory for Polrtrcal Scient~sts
(Princeton, N . J.: Princeton University Press, 1994) and Martin J. Osborne and Ariel Rubenstein. A
Course in Game Theory (Cambridge, Mass.: M I T Press, 1994).
I4See, for example, L. G. Anderson. The Economics of Fishery Management (Baltimore: Johns
Hopkins University Press, 1977).

Public Goods

93

i t will fish more intensively) as long as marginal private benefits-the rents they can
capture-xceed
marginal private costs. Just as in the restaurant example, each fisherman will ignore the marginal costs that his behavior imposes on other fishermen. If
every fisherman is equally efficient, then his marginal private cost equals the average
cost for the fishermen as a group. The question o f h o w rent should be distributed is
usually one of the most contentious issues in public policy. For example, h o w should
the catch limits for salmon be divided among commercial, sport, and native fishermen? Nevertheless, from the perspective of economic efficiency, someone should
receive the scarcity rent, rather than allowing i t to be wasted.
Note that demand for a resource may be sufficiently low so that i t remains uncongested at zero price. Increases in demand, however, may push the resource from
a free good (SWI) to an open access good (SW2). Historically, free goods have been
fairly common in North America, including such resources as buffalo, forests,
aquifer water, and rangeland. Typically, however, the free goods eventually disappeared as demand increased and excess supply was eliminated.I5 Open access often
led t o rapid depletion and, in some cases, near destruction o f the resource before effective exclusion was achieved. For example, open access permitted destruction o f
the Michigan, Wisconsin, and Minnesota pine forests.I6 Nonexcludibility continues
to be at the heart of many water use problems in the western United 5tates.l7 When
animal populations are treated as open access resources, the final result o f overconsumption may be species extinction; such a result has already occurred with certain
birds, and other animals with valuable skin or fur.
Thus far we have not specified the meaning o f "feasible" exclusion. Many o f
the goods we have given as examples appear not to be inherently nonexcludible. Indeed, one of the most famous historical examples of an open access resourcesheep and cattle grazing on the medieval English pasture-was "solved," willy-nilly,
without overt government intervention, by the enclosure movement, which secured
property rights for estate holders. Similar enclosures appear to be currently taking
place on tribal, historically open-access, lands in parts o f Africa.
It is useful to dichotomize open access and common property problems into
those that are structural (where aspects of the goods preclude economically feasible
exclusion mechanisms) and those that are institutional (where economically efficient
exclusion mechanisms are feasible but the distribution o f property rights precludes
their implementation). The averaging o f restaurant bills, previously discussed, serves
as an excellent illustration of an institutional common property problem. W e can
imagine an obvious exclusion mechanism that w e know from experience is economi-

I5lnthe case o f the American buffalo, the opening o f the railroad facilitated the hunt~ngand transportation o f hides at much lower costs so that what had prev~ouslybeen a free good became an open access resource. For an economlc analysis, see John Hanner, "Government Response to the Buffalo Hide
Trade, 1871-1883." JournolofLaw and Economks, Vol. 24, no. 2.1981. pp. 239-71.
I6Andrew Dana and John Baden. "The N e w Resource Economics: Toward an ldeologlcal Synthesis," Policy Studies Journal, Vol. 14, no. 2, 1985, pp. 2 3 3 4 3 .
"8. Delworth Gardner, "lnst~tutronalImpediments to Efic~entWater Allocation,'' Polrcy Studres
Review. Vol. 5 , no. 2, 1985, pp. 353-63. See also William Blomquist and Elinor Ostrom. "Institutional
Capacity and the Resolution o f a Commons Dilemma." Policy Studies Review, Vol. 5 , no. 2, 1985, pp.
283-93.

Rationales for Public Policy: Market Failures

Chap. 5

cally feasible: separate bills for everyone. Institutional common property resource
problems are usually not fundamentally market failures. Rather, they are most often
due to the failure of government to allocate enforceable property rights.
Typically, the crucial factor in making a distinction between structural and institutional problems is whether or not the good displays spatial stationarity. Trees are
spatially stationary, salmon are not, and bodies o f water may or may not be. When
resources are spatially stationary, their ownership can be attached to the ownership
o f land. Owners of the land are usually able to monitor effectively all aspects o f their
property rights and, consequently, ensure exclusive use. Given exclusion, common
property resources become private resources that will be used in an economically efficient manner. Without spatial stationarity, ownership o f land is not a good proxy
for low monitoring costs and the viability o f enforcing exclusion. It does not necessarily follow that the open access or common property problem could not be dealt
with by some form o f private ownership, but it does suggest that ownership o f a defined piece of land or water will not be adequate to ensure exclusion. Allocating fishing rights to specific water acreage where the fish stock moves over considerable
distances, or associating the rights to oil extraction to ownership o f land when the oil
pool extends under a large number o f parcels, illustrates the difficulty o f creating effective property rights for nonstationary resources.
In summary, a stationary good may have common property resource characteristics simply because its ownership is not well defined, perhaps because o f the historical accident that at one time supply exceeded demand at zero price. Nonstationary goods generally require more complex policy interventions to achieve efficiency
because the linking o f property rights to land ownership will not serve as an effective
proxy for exclusive resource ownership o f the resource.

petitive markets are not relevant extemalties because buyers and sellers are engaging
voluntarily in exchange. We have already encountered a variety o f externalities in
our discussion o f public goods-private supply o f nonrivalrous goods by privileged
and intermediate groups (a positive externality) and the divergence between marginal private and marginal social cost in the use o f congested resources (a negative
externality). W e reserve the label externalityproblem for those situations in which the
good conveying the valued impact on nonconsentingparties is the by-~roducto f either the production or consumption o f some good.
As is the case with open access resources and ambient public goods, externality problems involve attenuated property rights because either the rights to exclusive
use are incompletely specified or the costs o f enforcing the rights are high relative to
the benefits. Secure and enforceable property rights often permit private transactions to eliminate the economic inefficiency associated with an externality by opening up the possibility for markets in the external effect. Indeed, one can think o f an
externality problem as a missing market. We will return to this point after discussing a
few examples.
Common examples o f negative externalities include the air and water pollution
generated by firms in their production activities, the cigarette smoke that nonsmokers must breathe in public places, and the unsightliness generated by a dilapidated
house in a well-kept neighborhood. Persons who suffer these externalities place different values on them. For instance, I may simply dislike the smell o f cigarette
smoke, but you may be allergic to it. Whereas I would be willing to pay only a small
cost to avoid sitting near a smoker-say, waiting an extra ten minutes for a restaurant table in the nonsmoking section-you
might be willing to pay considerably
more-say, waiting thirty minutes or leaving the restaurant altogether. Note that
we can think of placing a value on these externalities in the same way we do for the
goods we voluntarily consume.
Common examples o f positive externalities include vaccinations that reduce
everyone's risk of infectious disease and the benefits that neighbors receive from a
homeowner's flower garden and nicely painted house. A n important category o f
positive externality arises in the context o f communication networks. One more person connecting a fax machine to the existing network o f fax machines provides marginal social benefits that exceed marginal private benefits because everyone already
Such positive'
on the network has one more interface for potential c~mrnunication.'~
externalities are usually referred to as network, or adoption, externalities.
Externalities can arise in either production or consumption. Production externalities affect either firms (producer-to-producer externalities) or consumers
(producer-to-consumer externalities); consumption externalites may also affect the
activities o f firms (consumer-to-producer externalities) or those o f other consumers
(consumer-to-consumerexternalites). In this context, consumers as recipients o f externalities includes everyone in society. Table 5.1 provides simple examples o f each
type o f externality. (Keep in mind that sometimes the same activity may constitute a
positive externality for some but a negative externality for others.) Classifying a situ- .ation that potentially involves an externality is often a good way to begin considering

Summary
Returning to Figure 5.2, we summarize the efficiency implications o f the various
types o f market failures involving public goods. T o reprise, the major problem with toll
goods (NE quadrant-nonrivalry, excludibility) is underconsumption arising from economically inefficient pricing rather than a lack o f supply per se. Congestion usually further complicates these problems by introducing the need for variable pricing to achieve
eficiency. In the case o f pure and ambient public goods (SE quadrant-nonrivalry,
nonexcludibility), the pervasiveness o f free riding generally leads to no market supply at
all. In specific circumstances (a privileged or intermediate group in which one or a few
persons account for a large fraction of demand), however, there may be some, and
perhaps even nearly efficient, market supply. In the case o f open access resources
(SW quadrant-rivalry,
nonexcludibility), inefficiency results because individuals do
not equate marginal social costs with marginal benefits, but rather marginal private
costs with marginal benefits; hence they overconsume (and overinvest in consumption
capacity) and underinvest in preserving open access resources.

EXTERNALITIES
'

'

A n externality I S any valued impact (positive or negative) resulting from any action
(whether related to production or consumption) that affects someone who did not
fully consent to it through participation in voluntary exchange. Price changes in com-

I8Foran overview o f network externalities, Michael L. Katz and Carl Shapiro. "Systems Competi
tion and Network Effects." JournalofEconomrc Perspectives. Vol. 8. no. 2.1994. pp. 9S115.

Rationales for Public Policy: Market Failures

Chap. 5

cally feasible: separate bills for everyone. Institutional common property resource
problems are usually not fundamentally market failures. Rather, they are most often
due t o the failure ofgovernment to allocate enforceable property rights.
Typically, the crucial factor in making a distinction between structural and institutional problems is whether or not the good displays spatial stationarity. Trees are
spatially stationary, salmon are not, and bodies o f water may or may not be. When
resources are spatially stationary, their ownership can be attached t o the ownership
o f land. Owners of the land are usually able t o monitor effectively all aspects o f their
property rights and, consequently, ensure exclusive use. Given exclusion, common
property resources become private resources that will be used in an economically efficient manner. Without spatial stationarity, ownership o f land is not a good proxy
for l o w monitoring costs and the viability o f enforcing exclusion. It does not necessarily follow that the open access or common property problem could not be dealt
w i t h by some form o f private ownership, but it does suggest that ownership o f a defined piece of land or water will not be adequate t o ensure exclusion. Allocating fishing rights t o specific water acreage where the fish stock moves over considerable
distances, or associating the rights t o oil extraction t o ownership o f land when the oil
pool extends under a large number o f parcels, illustrates the difficulty o f creating effective property rights for nonstationary resources.
In summary, a stationary good may have common property resource characteristics simply because its ownership is not well defined, perhaps because o f the historical accident that at one time supply exceeded demand at zero price. Nonstation-.
ary goods generally require more complex policy interventions t o achieve efficiency
because the linking o f property rights t o land ownership will not serve as an effective
proxy for exclusive resource ownership o f the resource.

Summary
Returning t o Figure 5.2, w e summarize the efficiency implications o f the various
types o f market failures involving public goods. T o reprise, the major problem with toll
goods ( N E quadrant--nonrivalry, excludibility) is underconsumption arising from economically inefficient pricing rather than a lack o f supply per se. Congestion usually further complicates these problems by introducing the need for variable pricing t o achieve
efficiency. In the case of pure and ambient public goods (SE quadrant-nonrivalry,
nonexcludibility), the pervasiveness o f free riding generally leads t o no market supply at
all. In specific circumstances (a privileged or intermediate group in which one or a few
persons account for a large fraction o f demand), however, there may be some, and
perhaps even nearly efficient, market supply. In the case o f open access resources
( S W quadrant-rivalry,
nonexcludibility), inefficiency results because individuals do
not equate marginal social costs with marginal benefits, but rather marginal private
costs with marginal benefits; hence they overconsume (and overinvest in consumption
capacity) and underinvest in preserving open access resources.

EXTERNALITIES
'

1
'

A n externality is any valued impact (positive or negative) resulting from any action
(whether related t o production or consumption) that affects someone w h o did not
fully consent t o i t through participation in voluntary exchange. Price changes in com-

Externalities

95

petitive markets are not relevant externalties because buyers and sellers are engaging
voluntarily in exchange. W e have already encountered a variety o f externalities in
our discussion o f public goods-private supply o f nonrivalrous goods by privileged
and intermediate groups (a positive externality) and the divergence between marginal private and marginal social cost in the use o f congested resources (a negative
externality). W e reserve the label externalityproblem for those situations in which the
good conveying the valued impact on nonconsenting parties is the by-product o f either the production or consumption o f some good.
As is the case w i t h open access resources and ambient public goods, externality problems involve attenuated property rights because either the rights t o exclusive
use are incompletely specified or the costs o f enforcing the rights are high relative to
the benefits. Secure and enforceable property rights often permit private transactions t o eliminate the economic inefficiency associated with an externality by opening up the possibility for markets in the external effect. Indeed, one can think of an
externality problem as a missing market. W e will return t o this point after discussing a
few examples.
Common examples of negative externalities include the air and water pollution
generated by firms in their production activities, the cigarette smoke that nonsmokers must breathe in public places, and the unsightliness generated by a dilapidated
house in a well-kept neighborhood. Persons w h o suffer these externalities place different values on them. For instance, I may simply dislike the smell o f cigarette
smoke, but you may be allergic t o it. Whereas I would be willing t o pay only a small
cost t o avoid sitting near a smoker-say, waiting an extra ten minutes for a restaurant table in the nonsmoking section-you
might be willing t o pay considerably
m o r e s a y , waiting thirty minutes or leaving the restaurant altogether. Note that
we can think o f placing a value on these externalities in the same way we do for the
goods w e voluntarily consume.
Common examples of positive externalities include vaccinations that reduce
everyone's risk o f infectious disease and the benefits that neighbors receive from a
homeowner's flower garden and nicely painted house. An important category o f
positive externality arises in the context o f communication networks. One more person connecting a fax machine t o the existing network o f fax machines provides marginal social benefits that exceed marginal private benefits because everyone already
'~
positive^
on the network has one more interface for potential c o m m ~ n i c a t i o n .Such
externalities are usually referred to as network, or adoption, externalities.
Externalities can arise in either production or consumption. Production externalities affect either firms (producer-to-producer externalities) or consumers
(producer-to-consumer externalities); consumption externalites may also affect the
activities o f firms (consumer-to-producer externalities) or those o f other consumers
(consumer-to-consumer externalites). In this context, consumers as recipients o f externalities includes everyone in society. Table 5.1 provides simple examples o f each
type ofexternality. (Keep in mind that sometimes the same activity may constitute a
positive externality for some but a negative externality for others.) Classifying a situ- ~ation that potentially involves an externality is often a good way t o begin considering

''For an ovewiew o f network externalities, Michael L. Katz and Carl Shapiro. "Systems Competition and Network Effects," Journal of Economlc Perspectjves, Vol. 8, no. 2, 1994, pp. 93-1 15.

96

Rationales for Public Policy: Market Failures

'

Chap. 5

97

Externalities

Table 5.1 Examples of Externalities

Positive

Negative

ProducertoProducer

Recreational facilities attracting

people who give custom to nearby businesses

Toxic chemical pollution harming

downstream commercial fishing

ProducertoConsumer

Private timber forests providing

scenic benefits to nature lovers

Air pollution from factories

harming lungs of people living
nearby

ConsumertoConsumer

Immunizationby persons against

contagious disease helping reduce risk to others

Cigarette smoke from one

person reducing enjoyment
of meal by another

Unsolicited letters from consumers

providing information on product
quality

Game hunters disturbing

domestic farm animals

MPC

ConsumertoProducer

its efficiency implications, distributional impacts, and, ryost importantly, possible

remedies.

Efficiency Losses of Negative and Positive Externalities

Figure 5.7 illustrates the resource allocation effects o f a negative externality in
production. In the presence o f a negative externality, firms will produce too much o f
the private good that generates the externality. The market supply schedule for the
private good, MPC, indicates the marginal costs borne directly by the firms producing it. For example, i f the private good were electricity, MPC would represent the
marginal amounts firms have to pay for the coal, labor, and other things that show
up in their ledgers. But M P C does not reflect the negative impacts o f the pollution
that results from burning the coal. If somehow we could find out how much each
person in society would be willing to pay to avoid the pollution at each output level,
then we could add these amounts to the marginal costs actually seen by the firms to
derive a supply schedule that reflected total social marginal costs. W e represent this
more ~nclusivesupply schedule as MSC.
Economic efficiency requires that social marginal benefits and social marginal
costs be equal at the selected output level-this occurs at quantity Q, where marg~nalsoc~alcost (MSC) and demand (D)intersect. But because firms do not consider
the external costs o f their output, they choose output level Q, at the intersection o f
M P C and D. Relative to output level Qo, consumers o f the private good being produced gain surplus equal to area PocaPe(because they get quantity Q, at price P,
rather than quantity Qo at price Po) and producers lose surplus equal to area PocdPe
minus area abd (the first area captures the effect o f the lower price, the second the
effect o f greater output). Those who bear the external costs, however, suffer a loss
given by the area abce (the area between the market and social supply curves over
the output difference-remember, the vertical distance between the supply curves

Social Surplus at Qe relative to Q0

Consumer surplus from private good:
Producer surplus of firms producing private good:
Loses to third parties bearing externality:
Social surplus:

larger by PocaPe
smaller by (PocdP, - abd)
larger by abce
smaller by ace

Figure 5.7 Overproduction with a Negative Externality

represents the external marginal cost). The net loss in social surplus is the area o f triangle ace, the algebraic sum o f the surplus differences for the consumers and producers ofthe private good and the bearers o f the externality. (This net social surplus loss
is simply the deadweight loss due to the overproduction-the area between MSC
and D from ,Q
, to Q,.) In other words, Pareto efficiency requires a reduction in output from the equilibrium level in the market (Q,) to the level at which marginal social
costs equal marginal social benefits (Qo).
Turning to positive externalities, we can think o f the private good generating
benefits that cannot be captured by the producer. For example, firms that plant trees
for future harvesting may provide a scenic benefit for which they receive no compensation. In Figure 5.8 we illustrate the demand schedule for the private good
(trees for future harvest) as D, which also gives the marginal private benefits (MPB).
At each forest size, however, the social marginal benefit schedule would equal the
market demand plus the amounts that viewers would be willing to pay to have the
forest expanded by another unit. MSB labels the social marginal benefit schedule,
which includes both the private and external marginal benefits. Market equilibrium

Rationales for Public Policy: Market Failures

Table 5.1

97

Externalities

Chap. 5

Examples o f Externalities

ProducerProker

Positive

Negative

Recreational facilities attracting

people who give custom to nearby businesses

Toxic chemical pollution harming

downstream commercial fishing

Private timber forests providing

scenic benefits to nature lovers

Air pollution from factories

harming lungs of people living
nearby

Consumer
contagious disease
ConsumertoProducer

of meal by another
Unsolicited letters from consumers
providing information on product
quality

Game hunters disturbing

domestic farm animals

its efficiency implications, distributional impacts, and, r?ost importantly, possible

remedies.

Efficiency Losses of Negative and Positive Externalities

Figure 5.7 illustrates the resource allocation effects o f a negative externality in
production. In the presence o f a negative externality, firms will produce too much o f
the private good that generates the externality. The market supply schedule for the
private good, MPC, indicates the marginal costs borne directly by the firms producing it. For example, i f the private good were electricity, MPC would represent the
marginal amounts firms have t o pay for the coal, labor, and other things that show
up in their ledgers. But MPC does not reflect the negative impacts o f the pollution
that results from burning the coal. If somehow w e could find out h o w much each
person in society would be willing t o pay t o avoid the pollution at each output level,
then we could add these amounts t o the marginal costs actually seen by the firms t o
derive a supply schedule that reflected total social marginal costs. W e represent this
more inclusive supply schedule as MSC.
Economic efficiency requires that social marginal benefits and social marginal
costs be equal at the selected output level-this occurs at quantity Qo where marglnal social cost (MSC) and demand (D) intersect. B u t because firms do not consider
the external costs o f their output, they choose output level Qe at the intersection o f
MPC and D. Relative t o output level Qo, consumers o f the private good being produced gain surplus equal t o area PocaPe(because they get quantity Q, at price P,
rather than quantity Qo at price Po) and producers lose surplus equal t o area PocdPe
minus area abd (the first area captures the effect o f the lower price, the second the
effect o f greater output). Those w h o bear the external costs, however, suffer a loss
given by the area abce (the area between the market and social supply curves over
the output difference-remember, the vertical distance between the supply curves

Social Surplus at Qe relative to QO

Consumer surplus from private good:
Producer surplus of firms producing private good:
Loses to third parties bearing externality:
Social surplus:

larger by PocaPe
smaller by (PocdP, - a b d )
larger by abce
smaller by ace

Figure 5.7 Overproduction with a Negative Externality

represents the external marginal cost). T h e net loss in social surplus is the area o f triangle ace, the algebraic sum o f the surplus differences for the consumers and producers o f the private good and the bearers o f the externality. (This net social surplus loss
is simply the deadweight loss due to the overproduction-the
area between M S C
and D from Qot o Q,.) In other words, Pareto efficiency requires a reduction in output from the equilibrium level in the market (Q,) t o the level at which marginal social
costs equal marginal social benefits (Qo).
Turning t o positive externalities, w e can think o f the private good generating
benefits that cannot be captured by the producer. For example, firms that plant trees
for future harvesting may provide a scenic benefit for which they receive no compensation. In Figure 5.8 w e illustrate the demand schedule for the private good
(trees for future harvest) as D, which also gives the marginal private benefits (MPB).
At each forest size, however, the social marginal benefit schedule would equal the
market demand plus the amounts that viewers would be willing to pay t o have the
forest expanded by another unit. M S B labels the social marginal benefit schedule,
which includes both the private and external marginal benefits. Market equilibrium

98

Rationales for Public Policy: Market Failures

Chap. 5

99

Externalities

Social Surplus at Qerelative to Q0

Consumer surplus: smaller by (acd - PocbPe)
Producer surplus: smaller by PoabPe
Social surplus:
smaller by abd
Figure 5.8

Underproduction with a Positive Externality

results at output level Q, where the market demand schedule and the supply schedule intersect. But i f output were raised t o Qo, consumer surplus would increase by
area acd (resulting from increased consumption from Q, t o Qo) minus area PocbPe
(due to the price rise from P, to Po), and producer surplus would increase by area
PoabP, . The net increase in social surplus, therefore, would be area abd. Again, we
see that in the case of an externality w e can find a reallocation that increases soc~al
surplus and thereby offers the possibility o f an increase in efficiency.

in a seminal article on the externality problem.'9 H e argued that in situations in

which property rights are clearly defined and costless to enforce, costless bargaining
among participants will lead to an economically efficient level o f the external effect.
Of course, the distribution o f costs and benefits resulting from the bargaining depends on w h o has the property rights.
Before exploring the issue further, we must stress a very restrictive assumption o f Coase's model that limits its applicability to many actual externality situations. Namely, Coase assumes zero transaction costs in the exercise o f property
rights.'' In many real-world situations, transaction costs are high usually because
those producing and experiencing the externality are numerous. W i t h large numbers
o f actors, the bargaining that produces the Coasian'outcome becomes impossible
because o f the high costs o f coordination in the face o f opportunities for a free ride.
Nevertheless, Coase's insight is valuable. H e pointed out that in the case o f
smallnumbers (he assumes one source and one recipient o f the externality), the allocation o f property rights alone would lead t o a negotiated (that is, private) outcome
that is economically efficient. Assuming that individuals make decisions based on the
dollar value rather than the utility value o f external effects, efficiency results
whether a complete property right is given t o the externality generator (one bears no
liability for damages caused by one's externality) o r t o the recipient o f the externality
(one bears full liability for damages caused by one's externality). Either rule should
lead to a bargain being reached at the same level o f externality; only the distribution
o f wealth will vary, depending on which rule has force. Assuming that individuals
maximize utility rather than wealth, however, opens the possibility for different allocations under different property rights assignments.21
A moment's thought should suggest why large numbers will make a Coasian
solution unlikely. Bargaining would have to involve many parties, some o f whom
would have an incentive t o engage in strategic behavior. For example, in the case o f
a polluter with no liability for damages, those experiencing the pollution may fear free
riding by others. In addition, firms may engage in opportunistic behavior, threatening
to generate more pollution to induce payments. Under full liability, individuals would
have an incentive to overstate the harm they suffer. Other things equal, we expect
that the greater the number o f parties experiencing the externality, the greater will
be the costs o f monitoring damage claims.
Nevertheless, private cooperation appears effective in some situations. For instance, neighborhood associations sometimes do agree on mutually restrictive
covenants, and individual neighbors occasionally do reach contractual agreements on
such matters as light easements (which deal with the externalities o f the shadows
cast by buildings).

Market Responses to Externalities

19~onald
Coase. "The Problem o f Soclal Cost." Journalof Law and Economics. Vol. 3, no. 1,
pp.

Will the market always fail to provide an efficient output level in the presence
o f externalities? Just as pure public goods will sometimes be provided at efficient, or
nearly efficient, levels through voluntary private agreements within intermediate
groups, so too may private actions counter the inefficiency associated with externalities. The relevance of such private responses was first pointed out by Ronald Coase

1960,

1-44,
'O~ora conceptual discussion o f transaction costs that takes account o f barga~ning,see Douglas

D. Heckathorn and Steven M . Maser, "Bargaining and the Sources o f Transaction Costs: The Case o f
Government Regulation." Journalof Law, Economics, and Organization. Vol. 3 , no. I . 1987,pp. 69-98.

o or a discussion o f this point and a general overview o f Coase, see Thrainn Eggertsson. Eco-

nomic Behawor and institutions (New York: Cambridge University Press.

1990). pp. 101-1 10.

Rationales for Public Policy: Market Failures

Chap. 5

Social Surplus at Q, relative to Q0

Consumer surplus: smaller by (acd - PocbP,)
Producer surplus: smaller by PoabP,
Social surplus:
smaller by abd
Figure 5.8

Underproduction with a Positive Externality

results at output level Q, where the market demand schedule and the supply schedule intersect. But if output were raised to Qo, consumer surplus would increase by
area acd (resulting from increased consumption from Q, t o Qo) minus area PocbP,
(due to the price rise from P, to Po), and producer surplus would increase by area
PoabPe. The net increase in social surplus, therefore, would be area abd. Again, we
see that in the case of an externality w e can find a reallocation that increases social
surplus and thereby offers the possibility o f an increase in efficiency.

Market Responses to Externalities

Will the market always fail to provide an efficient output level in the presence
o f externalities? Just as pure public goods will sometimes be provided at effic~ent,or
nearly efficient, levels through voluntary private agreements within intermediate
groups, so too may private actions counter the inefficiency associated with externalities. The relevance o f such private responses was first pointed out by Ronald Coase

Externalities

99

in a seminal article on the externality problem.'9 H e argued that in situations in

which property rights are clearly defined and costless to enforce, costless bargaining
among participants will lead t o an economically efficient level o f the external effect.
Of course, the distribution o f costs and benefits resulting from the bargaining depends on who has the property rights.
Before exploring the issue further, w e must stress a very restrictive assumption o f Coase's model that limits its applicability t o many actual externality situations. Namely, Coase assumes zero transaction costs in t h e exercise o f property
rights.'' In many real-world situations, transaction 'costs are high usually because
those producing and experiencing the externality are numerous. With large numbers
o f actors, the bargaining that produces the Coasian'outcome becomes impossible
because o f the high costs o f coordination in the face o f opportunities for a free ride.
Nevertheless, Coase's insight is valuable. H e pointed out that in the case of
smallnumbers (he assumes one source and one recipient o f the externality), the allocation o f property rights alone would lead t o a negotiated (that is, private) outcome
that is economically efficient. Assuming that individuals make decisions based on the
dollar value rather than the utility value o f external effects, efficiency results
whether a complete property right is given t o the externality generator (one bears no
liability for damages caused by one's externality) or: t o the recipient o f the externality
(one bears full liability for damages caused by one's externality). Either rule should
lead to a bargain being reached at the same level o f externality; only the distribution
o f wealth will vary, depending on which rule has force. Assuming that individuals
maximize utility rather than wealth, however, opens the possibility for different allocations under different property rights assignments."
A moment's thought should suggest why large numbers will make a Coasian
solution unlikely. Bargaining would have to involve many parties, some o f whom
would have an incentive t o engage in strategic behavior. For example, in the case of
a polluter with no liability for damages, those experiencing the pollution may fear free
riding by others. In addition, firms may engage in opportunistic behavior, threatening
to generate more pollution to induce payments. Under full liability, individuals would
have an incentwe t o overstate the harm they suffer. Other things equal, w e expect
that the greater the number o f parties experiencing the externality, the greater will
be the costs o f monitoring damage claims.
Nevertheless, private cooperation appears effective in some situations. For instance, neighborhood associations sometimes do agree on mutually restrictive
covenants, and individual neighbors occasionally do reach contractual agreements on
such matters as light easements (which deal with the externalities o f the shadows
cast by buildings).

19RonaldCoase. "The Problem of Social Cost," Journal of Low and Economics, Vol. 3 , no. 1, 1960,
pp. 144.
''For a conceptual discussion o f transaction costs that takes account o f bargaining, see Douglas
D. Heckathorn and Steven M . Maser. "Bargaining and the Sources of Transaction Costs: The Case of
Government Regulation," JournolofLow. Economics. and Orgonizotion. Vol. 3. no. 1, 1987, pp. 69-98.
"For a d~scussiono f this point and a general overview o f Coase, see Thrainn Eggertsson. Economic Behmior and institutions (New York: Cambridge University Press. 1990). pp. 101-1 10.

100
,

Rationales for Public Policy: Market Failures

Chap. 5

.
Moreoever, there is an important case where Coase-like solutions do arise,
even w i t h large numbers o f parties-namely, where (I) property rights become implicitly established by usage; (2) the value o f the externality (whether positive or negative) is captured by (more technically, "capitalized into") land values; (3) considerable time has passed such that the initial stock o f external parties has "rolled over";
and (4) externality levels remain stable. The relevance o f these conditions can best
be explained w i t h an example. Suppose that a factory has been polluting the surrounding area for many years without anyone challenging its owners' rights t o do so.
It is probable that the pollution will result in lower property values." Residents who
bought before the pollution was anticipated will have t o sell their houses for lessreflecting the impact o f the pollution. N e w homeowners, however, will not be bearing any Pareto-relevant externality, because the negative impact o f the pollution will
be capitalized into house prices. The lower prices o f the houses will reflect the market's (negative) valuation o f the pollution. In other words, through the house price i t
is possible t o get a proxy dollar measure o f the disutility o f pollution. Notice that a
second generation o f homeowners (that is, those who bought houses after the pollution was known and capitalized into prices) would receive a bonus i f the existing allocation o f property rights were changed so that the factory had t o compensate current homeowners for existing levels o f pollution. Of course, if there are unexpected
changes in the level o f pollution (or new information about the harmful impacts o f
the pollution-see our discussion o f information asymmetry below), there will be in
effect new (either positive or negative) impacts; in these situations, considerable argument is likely t o occur over w h o has right, t o compensation for the changes.

101

Natural Monopoly

manded), then w e say that demand is inelastic and an increase in price will increase
total revenue. A good is unlikely t o have inelastic demand i f there are other products
that, while not exactly the same, are close substitutes. In such circumstances, the
availability o f substitutes greatly limits the economic inefficiency associated with natural monopoly. For example, although local cable television markets have the costand-demand characteristics o f natural monopolies, many substitute products, including conventional over-the-air television, video tape recorders (and, arguably, all other
forms o f entertainment) may prevent the cable television companies from realizing
large monopoly rents in their local markets.
W e should also keep in mind that, although w e stated the basic definition o f
natural monopoly in static terms, markets in the real world are dynamic. Technological change may lead t o different cost characteristics or high prices may bring on close
or even superior substitutes. The result may be the elimination o f natural monopoly
or the supplanting of one natural monopoly technology by another. For example, direct satellite is a direct substitute for cable television, raising the possibility o f greater
competition in the distr~butiono f television programming.

Allocative Inefficiency under Natural Monopoly

W e show a cost structure leading t o natural monopoly in Figure 5.9. Marginal
cost (MC) is shown as constant over the range o f output. Fixed cost (FC), which is
not shown on the figure, must be incurred before any output can be supplied. Because o f the fixed cost, average cost (AC) starts higher than marginal cost and falls
toward it as the output level increases. ( A C = FC/Q M C , where Q is the output
level.) Although marginal cost is shown as constant, the same analysis would apply
even i f marginal cost were rising or falling, provided that it remains small relative t o
fixed cost.
Figure 5.9 shows the divergence between the firm's profit-maximizing behavior and economic efficiency. Let us first consider the economically efficient price and
output. Efficiency requires, as w e discussed in the previous chapter, that price be set
equal t o marginal cost (P = M C ) . Because marginal cost is below average cost, the
economically efficient output level Qo results in the firm suffering a loss equal to FC.
Obviously, the firm would not choose this output level on its own. Instead, it would
maximize profits by selecting output level Q
,,
which equates marginal revenue t o
,, the market price will be P
, and profmarginal cost (MR = MC). At output level Q
its equal t o the area of rectangle P,,,cbPo - FC will result. Relative t o output level Qo,
consumer surplus is lower by the area o f PmcaPo,butthe profit o f the firm is larger by
PmcbPo.The net loss In social surplus due t o the underproduction equals the area o f
the shaded triangle abc, the deadweight loss t o consumers. (As noted in our discussion o f Figure 4.5, unrts o f forgone output would have offered marginal benefits in
excess of marginal costs-a total loss equal t o the area between the marginal benefit, or demand schedule, and the marginal cost curve. Again, in Figure 5.9, thrs loss is
the area o f triangle abc.)

NATURAL MONOPOLY
Natural monopoly occurs when average cost declines over the relevant range o f demand. Note that this definition is in terms o f both cost and demand conditions. In
the case o f natural monopoly, a single firm can produce the output at lower cost
than any other market arrangement, including competition.
While the cost-and-demand conditions establish the existence o f a natural monopoly situation, the price elasticity of demand determines whether or not the natural
monopoly has important implications for public policy. The price elasticity o f demand
measures h o w responsive consumers are t o price changes. Specifically, the price
elasticity o f demand is defined as the percentage change in the quantity demanded
(measured in percentage points) that results from a one-percentage point change in
price.23 If the absolute value o f the price elasticity o f demand is less than one (a onepercent change in price leads t o less than a one-percent reduction in the quantity de-

"Indeed, changes in property values provide a basis for empirically est~matingthe social costs o f
externalities. For example, with respect t o air pollution, see V. Kerry Sm~thand Ju-Chin Huang, "Can
Markets Value Air Quality? A Meta-Analysis o f Hedonic Property Value Models." Journal of Political
Economy, Vol. 103, no. 1, 1995, p p 209-227.
Z3Mathematically,if the demand schedule is continuous, the price elasticity o f demand at some
quantity equals the slope o f the demand schedule at that quant~tytimes the ratlo of quant~tyto price. For
example, with h e a r demand schedule, Q = a - bP, the slope o f the demand schedule (the derivative

dQ/dP) is -b. Therefore, the price elasticity o f demand is e = - bP/Q. Note that the elasticity o f a linear
demand schedule varies with quantity

Rationales for Public Policy: Market Failures

Chap. 5

Moreoever, there is an important case where Coase-like solutions do arise,

even w i t h large numbers o f parties-namely, where (1) property rights become implicitly established by usage; (2) the value of the externality (whether positive or negative) is captured by (more technically, "capitalized into") land values; (3) considerable time has passed such that the initial stock o f external parties has "rolled over";
and (4) externality levels remain stable. The relevance o f these conditions can best
be explained w i t h an example. Suppose that a factory has been polluting the surrounding area for many years without anyone challenging its owners' rights t o do so.
It is probable that the pollution will result in lower property values.22Residents who
bought before the pollution was anticipated will have to sell their houses for lessreflecting the impact o f the pollution. N e w homeowners, however, will not be bearing any Pareto-relevant externality, because the negative impact o f the pollution will
be capitalized into house prices. The lower prices o f the houses will reflect the market's (negative) valuation o f the pollution. In other words, through the house price it
i s possible t o get a proxy dollar measure o f the disutility o f pollution. Notice that a
second generation o f homeowners (that is, those who bought houses after the pollution was known and capitalized into prices) would receive a bonus if the existing allocation o f property rights were changed so that the factory had t o compensate current homeowners for existing levels o f pollution. Of course, if there are unexpected
changes in the level of pollution (or new information about the harmful impacts o f
the pollution-see our discussion o f information asymmGtry below), there will be in
effect new (either positive or negative) impacts; in these situations, considerable argument is likely t o occur over who has rrght, t o compensation for the changes.

NATURAL MONOPOLY
Natural monopoly occurs when average cost declines over the relevant range of demand. Note that this definition is in terms of both cost and demand conditions. In
the case o f natural monopoly, a single firm can produce the output at lower cost
than any other market arrangement, including competition.
While the cost-and-demand conditions establish the existence o f a natural monopoly situation, the prlce elasticity of demand determines whether or not the natural
monopoly has important implications for public policy. The price elasticity o f demand
measures h o w responsive consumers are t o price changes. Specifically, the price
elasticity o f demand is defined as the percentage change in the quantity demanded
(measured in percentage points) that results from a one-percentage point change in
price.23If the absolute value o f the price elasticity o f demand is less than one (a onepercent change in price leads t o less than a one-percent reduction in the quantity de-

221ndeed,changes In property values provlde a bass for empirically estimating the social costs of
externaltties. For example, with respect to air pollut~on,see V. Kerry Smith and Ju-Chin Huang, "Can
Markets Value Air Qualrty? A Meta-Analys~so f Hedonic Property Value Models," Journal of Political
Economy. Vol. 103, no. 1. 1995, pp 209-227.
23Mathematlcally, if the demand schedule IS continuous, the price elasticity o f demand at some
quantity equals the slope o f the demand schedule at that quantity tlmes the ratlo of quantity to price. For
example, wtth linear demand schedule, Q = a - bP, the slope o f the demand schedule (the derivative

Natural Monopoly
manded), then w e say that demand is inelastic and an increase in price will increase
total revenue. A good is unlikely t o have inelastic demand i f there are other products
that, while not exactly the same, are close substitutes. In such circumstances, the
availability o f substitutes greatly limits the economic inefficiency associated with natural monopoly. For example, although local cable television markets have the costand-demand characteristics o f natural monopolies, many substitute products, including conventional over-the-air television, video tape recorders (and, arguably, all other
forms o f entertainment) may prevent the cable television companies from reallzing
large monopoly rents in their local markets.
W e should also keep in mind that, although w e stated the basic definition o f
natural monopoly in static terms, markets in the real world are dynamic. Technological change may lead t o different cost characteristics or high prices may bring on close
or even superior substitutes. The result may be the elimination o f natural monopoly
or the supplanting o f one natural monopoly technology by another. For example, direct satellite is a direct substitute for cable television, raising the possibility o f greater
competition in the distribution o f television programming.

Allocative Inefficiency under Natural Monopoly

W e show a cost structure leading t o natural monopoly in F~gure5.9. Marginal
cost (MC) is shown as constant over the range o f output. Fixed cost (FC), which is
not shown on the figure, must be incurred before any output can be supplied. Because o f the fixed cost, average cost (AC) starts higher than marginal cost and falls
toward i t as the output level increases. ( A C = FC/Q + M C , where Q is the output
level.) Although marginal cost is shown as constant, the same analysis would apply
even i f marginal cost were using or falling, provided that it remains small relative t o
fixed cost.
Figure 5.9 shows the divergence between the firm's profit-maximizing behavior and economic efficiency. Let us first consider the economically efficient price and
output. Efficiency requires, as we discussed in the previous chapter, that price be set
equal t o marginal cost (P = M C ) . Because marginal cost is below average cost, the
economically efficient output level Qo results in the firm suffering a loss equal to FC.
Obviously, the firm would not choose this output level on its own. Instead, i t would
,,
which equates marginal revenue t o
maximize profits by selecting output level Q
, the market price will be P, and profmarginal cost (MR = M C ) . At output level Q
its equal t o the area o f rectangle Pmc6P0- FC will result. Relative t o output level Qo,
consumer surplus is lower by the area o f P,caPo, but the profit o f the firm is larger by
Pmc6Po.The net loss in social surplus due t o the underproduction equals the area o f
the shaded triangle abc, the deadweight loss to consumers. (As noted in our discussion o f Figure 4.5, units o f forgone output would have offe'red marginal benefits in
excess o f marginal costs-a total loss equal to the area between the marginal benefit, or demand schedule, and the marginal cost curve. Again, in Figure 5.9, t h ~ sloss is
the area o f triangle abc.)

IS -6. Therefore, the prlce elasticity o f demand is e = - bP/Q. Note that the elasticity o f a linear
demand schedule vanes wlth quantity

dQ/dP)

Rationales for Public Policy: Market Failures

Chap. 5

103

Natural Monopoly

duced QAC in Figure 5.9. Clearly, under these circumstances, the firm can survive
because its costs are n o w being just covered. (Note that FC equals PAcefPo.) Although the deadweight loss under average cost pricing is much lower than under
monopoly pricing (area aef versus area abc ), i t is not eliminated. Therefore, average
cost pricing represents a compromise t o the natural monopoly dilemma.
Restraints W h e n

Consumer surplus:
Total revenue:
Total cost:
Producer surplus:
Social surplus:

Monopoly Pricing
(P,?,)
P,,,cd
P,i,cQ,,O
FC + PobQltiO
P,,,cbPo - FC
Podcb - FC

Efficient Pricing
Average Cost Pricing
(Po)
(p~c)
Pond
PoflQoo
PAC~QACO
FC + PonQoO
FC + P o f Q ~ c o
- FC
0
Poad - FC

Markets A r e Contestable

W e can imagine circumstances in which the natural monopoly firm might, in

fact, be forced t o price at average cost because o f the threat o f competition from potential entrants. The crucial requirement is that entry to, and exit from, the industry
be relatively easy. Whether or not the firm has in-place capital that has no alternative use, capital whose costs are sunk, usually determines the viability o f potential
entry and exit. If the established natural monopoly has a large stock o f productive
capital that cannot be sold for use in other industries, it will be difficult for other
firms t o compete because they must first incur cost to catch up with the established
firm's capital advantage. The in-place capital serves as a barrier t o entry; the greater
the replacement cost o f such capital, the higher the barrier to entry. For example,
once a petroleum pipeline is built between t w o cities, its scrap value is likely t o be
substantially less than the costs a potential competitor would face in building a second pipeline. The greater the difference is, the greater the ability is o f the established
firm t o fight o f f an entry attempt with temporarily lower prices. Of course, keep in
mind that the ability o f the firm t o charge above the marginal cost level will be influenced by the marginal costs o f alternative transportation modes, such as truck and
rail, that can serve as substitutes.
Much recent economic research considers industries with low barriers t o
entry and decreasing average costs, which, because o f the threat o f potential entry,
W e expect markets that are contestable t o exare said to be in contestable
hibit pricing closer t o the efficient level. One o f the most important debates in the literature arising from the contestable market framework concerns the empirical significance o f in-place capital as effective barriers t o entry.25Most natural monopolies
appear t o enjoy large advantages in in-place capital, raising the question o f whether
or not they should be viewed as being in contestable markets.

abc
Net social surplus loss of monopoly pricing relative to efficient pricing:
Net social surplus loss of average cost pricing relative to efficient pricing: aef

Figure 5.9 Social Surplus Loss from Natural Monopoly

74W~ll~am
J Baumol, John C Panzar, and Robert D Wllllg, Contestable Markets and the Theory of
Industry Structure (New York Harcourt Brace Jovanov~ch.1982), and Wllllam J Baurnol, "Contestable

Imagine that public policy forced the natural monopolist t o price at the economically efficient level (Po).T h e monopolist would suffer a loss o f FC. Consequently, the monopolist would go out o f business in the absence of a subsidy t o
offset the loss. Notice the dilemma presented by natural monopoly: Forcing the monopolist t o price efficiently drives it out o f business; allowing the monopolist t o set
price to maximize profits results in deadweight loss.
Briefly consider what would happen if the firm were forced by public policy to
price at average, rather than marginal, cost; that is, i f the firm priced at PACand pro-

Markets A n Uprlslng In the Theory o f Industry Structure," Amerrcan Economrc Rewew, Vol 72, no I,
1982, pp 1-15 For a dlscuss~ono f finperfect contestab~lrty.see Steven A Morrtson and Cllfford W~nston,
" Emplrlcal lmpllcat~onsand Tests o f the Contestablllty Hypothesis," Journal of Low and Economas, Vol
30, no 1, 1987 pp 53-66
25Forevidence that the U.S. Postal Service, for example, may have few natural monopoly characteristics, see Alan L. Sorkin, The Economics of the Postal Service (Lexington, Mass.: Lexington Books,
1980). Chapter 4; Leonard Waverman. "Pricing Principles: H o w Should Postal Rates Be Set?" in Perspectives on Postal Servlces Issues. Roger Sherman, ed.. (Washington. D.C.: American Enterprise Instit~lte,
1980), pp. 7-26.

R a t i o n a l e s f o r P u b l i c P o l i c y : M a r k e t Failures

Chap. 5

Natural Monopoly

103

duced QAC in Figure 5.9. Clearly, under these circumstances, the firm can survive
because its costs are now being just covered. (Note that FC equals PAcef Po.) Although the deadweight loss under average cost pricing is much lower than under
monopoly pricing (area aef versus area abc ), i t is not eliminated. Therefore, average
cost pricing represents a compromise t o the natural monopoly dilemma.

Restraints When Markets Are Contestable

Consumer surplus:
Total revenue:
Total cost:
Producer surplus:
Social surplus:

Monopoly Pricing
(Pllt)
P111cd
Prii~QiiiO
FC + PobQll,O
P,,,cbPo- FC
Podcb - FC

Efficient Pricing
Average Cost Pricing
(Po)
( p ~ ~ )
Poad
PoaQoo
PAC~QACO
FC + PonQoO
FC + P o f Q ~ c o
- FC
0
Poad - FC

W e can imagine circumstances in which the natural monopoly firm might, in

fact, be forced to price at average cost because o f the threat o f competition from potential entrants. The crucial requirement is that entry to, and exit from, the industry
be relatively easy. Whether or not the firm has in-place capital that has no alternative use, capital whose costs are sunk, usually determines the viability o f potential
entry and exit. If the established natural monopoly has a large stock o f productive
capital that cannot be sold for use in other industries, i t will be difficult for other
firms t o compete because they must first incur cost t o catch up with the established
firm's capital advantage. The in-place capital serves as a barrier t o entry; the greatei
the replacement cost o f such capital, the higher the barrier to entry. For example,
once a petroleum pipeline is built between t w o cities, its scrap value is likely t o be
substantially less than the costs a potential competitor would face in building a second pipeline. T h e greater the difference is, the greater the ability is o f the established
firm to fight o f f an entry attempt with temporarily lower prices. Of course, keep in
mind that the ability o f the firm t o charge above the marginal cost level will be influenced by the marginal costs o f alternative transportation modes, such as truck and
rail, that can serve as substitutes.
Much recent economic research considers industries with low barriers t o
entry and decreasing average costs, which, because o f the threat o f potential entry,
are said t o be in contestable
W e expect markets that are contestable to exhibit pricing closer t o the efficient level. One o f the most important debates in the literature arising from the contestable market framework concerns the empirical significance o f in-place capital as effective barriers to entry.25Most natural monopolies
appear to enjoy large advantages in in-place capital, raising the question o f whether
or not they should be viewed as being in contestable markets.

abc
Net social surplus loss of monopoly pricing relative to efficient pricing:
Net social surplus loss of average cost pricing relative to efficient pricing: aef

Figure 5.9 S o c i a l Surplus Loss f r o m N a t u r a l M o n o p o l y

'4William J. Baurnol, John C. Panzar, and Robert D. Willig. Contestable Markets and the Theory of
(New York: Harcourt Brace Jovanovich, 1982); and W~lliamJ. Baumol. "Contestable
Markets: A n Upris~ngin the Theory of lndustry Structure." American Economic Review. Vol. 72, no. I,
1982, pp. 1-15. For a discussion o f ,inperfect contestabilrty. see Steven A. Morrison and Clifford Winston,
"Empirical Implications and Tests o f the Contestability Hypothesis," Journal of Law and Economics, Vol.
30, no. I, 1987. pp. 53-66.
25Forevidence that the U.S. Postal Serv~ce,for example. may have few natul-almonopoly characteristics, see Alan L. Sorkin, The Economics of the Postal Service (Lexington. Mass.: Lexington Books.
1980). Chapter 4; Leonard Waverman, "Pricing Principles: H o w Should Postal Rates Be Set?" in Perspectives on Postal Services Issues. Roger Sherman, ed., (Washington, D.C.: American Enterprise Institute.
1980), pp. 7-26.
Industry Structure

Imagine that public policy forced the natural monopolist t o price at the economically efficient level (Po). The monopolist would suffer a loss o f FC. Consequently, the monopolist would go out o f business in the absence o f a subsidy t o
offset the loss. Notice the dilemma presented by natural monopoly: Forcing the monopolist t o price efficiently drives i t out o f business; allowing the monopolist to set
price t o maximize profits results in deadweight loss.
Briefly consider what would happen if the firm were forced by public policy t o
price at average, rather than marginal, cost; that is, i f the firm priced at PACand pro-

Rationales for Public Policy: Market Failures

Chap. 5

As w e have seen, the "naturalness" o f a monopoly is determined by decreasing

average cost over the relevant range o f output. In many situations, i t appears that
average cost declines over considerable ranges o f output, but then flattens out. In
other words, initial economies o f scale are exhausted as output increases. What happens i f demand shifts t o the right (for example, with population increases) such that
the demand curve intersects the flat portion o f the average cost curve? There may
be considerable room for competition under these circumstances. Figure 5.10 illustrates such a situation. I f demand is only D, (the "classic" natural monopoly), only
one firm can survive in the market. I f demand shifts outward beyond DZt o say, D3,
then t w o or more firms may be able t o survive because they can each operate on the
flat portion o f the average cost curve.
Simply because t w o firms could survive at D3 does not mean that t w o competing firms will actually emerge as demand expands. If the original natural monopoly
firm expands as demand moves out, it may be able t o forestall new entrants and cap-

Natural Monopoly

105

ture all o f the market. Nevertheless, we are again reminded o f the importance o f
looking beyond the static view.
When considering natural monopoly from a policy perspect~ve,w e often find
that legal and regulatory boundaries do not correspond to the boundaries delineating
natural monopoly goods. Most discussion of industrial issues tends t o be within the
framework o f product "sectors," such as the electric and the telephone industries.
Unfortunately, the economic boundaries of natural monopolies are not likely t o conform t o these neat sectoral boundaries. Historically, regulation has often not recognized this unpleasant fact. In addition, the existence, or extent, o f a natural monopoly can change as production technology or demand changes.
The telecommunications industry illustrates these definitional problems. In
1982, the U.S. Justice Department and ATGT agreed on a plan for the breakup o f
the corporation. The agreement called for a division into t w o parts: the part that
could be expected to be workably contestable and the part that had strong natural
monopoly elements. Both long-distance services (contestable) and equipment manufacture and supply (competitive) were largely deregulated, while local telephone exchanges were deemed t o be regional natural monopolies.2~imilarproblems w i t h
sectoral definitions have not been well recognized in other contexts, however. For
example, electricity generation and transmission have typically been treated as part
o f an electrical utility natural monopoly, although the evidence suggests that in many
circumstances only transmission has the required cost and demand characteristics t o
be considered a natural monopoly.27 (Running multiple sets of transmission lines
across the countryside would generally be inefficient; having multiple firms generating
.,electricity would not be.)
Another dimension, apart from the sectoral, where the problem o f defining
natural monopoly arises is the spatial. Over what spatial area does the natural monopoly exist? Almost all natural monopoly regulation corresponds t o existing city,
county, state, and federal boundaries; but the economic reality o f a natural rnonopoly knows no such bounds-it is purely an empirical question h o w far (spatially) the
natural monopoly extends. Again, the appropriate spatial boundaries o f a natural monopoly can change with changes in technology.

X-Inefficiency Resulting from Limited Competition

Thus far, w e have described the potential social costs o f natural monopoly.
whether it prices t o either maximize profits or cover cost, in terms o f deadweight
loss caused by allocational inefficiency. The social costs, however, may be larger be-

26~enneth
Robinson. "Max~rnlzingthe Public Benefits o f the AT&T Breakup," Journal of Policy
3. 1986, p p 572-97 For discussion o f technological change that
eroded the natural monopoly characteristics o f telephone services, see Irwin Madey, Telecommunications
America: Markets Wrthout Boundarres (Westport. Conn.: Quorum. 1984).
27 For the case against treatlng any aspects o f the electric ~ndustyas a natural monopoly, see
Analysrs and Management. Vol. 5. no.

Figure 5.10 Shifting Demand and Multiple Firm Survival (Source:

William G. Shepherd, The Economics of Industrial Organization
(Englewood Cliffs, N.J.: Prentice Hall, 1979), fig. 3.4, p. 59.)

Robert W. Poole. Jr., Unnatural Monopolies- The Case for Deregulating Public Utilities (Lexington. Mass.:

Lexington Books, 1985).

Rationales for Public Policy: Market Failures

Chap. 5

As we have seen, the "naturalness" o f a monopoly is determined by decreasing

average cost over the relevant range o f output. In many situations, i t appears that
average cost declines over considerable ranges o f output, but then flattens out. In
other words, initial economies o f scale are exhausted as output increases. What happens i f demand shifts to the right (for example, with population increases) such that
the demand curve intersects the flat portion o f the average cost curve? There may
be considerable room for competition under these circumstances. Figure 5.10 illustrates such a situation. I f demand is only D, (the "classic" natural monopoly), only
one firm can survive in the market. I f demand shifts outward beyond D2to say, D3,
then two or more firms may be able to survive because they can each operate on the
flat portion o f the average cost curve.
Simply because two firms could survive at D3does not mean that two competing firms will actually emerge as demand expands. If the original natural monopoly
firm expands as demand moves out, it may be able to forestall new entrants and cap-

Natural Monopoly

105

ture all o f the market. Nevertheless, we are again reminded o f the importance o f
looking beyond the static view.
When considering natural monopoly from a policy perspective, we often find
that legal and regulatory boundaries do not correspond to the boundaries delineating
natural monopoly goods. Most discuss~ono f industrial issues tends to be within the
framework of product "sectors," such as the electric and the tele~honeindustries.
Unfortunately, the economic boundaries of natural monopolies are not likely to conform to these neat sectoral boundaries. Historically, regulation has often not recognized this unpleasant fact. In addtion, the existence, or extent, o f a natural monopoly can change as production technology or demand changes.
The telecommunicat~ons~ndustryillustrates these definitional problems. In
1982, the U.S. Justlce Department and AT&T agreed on a plan for the breakup o f
the corporation. The agreement called for a division into two parts: the part that
could be expected to be workably contestable and the part that had strong natural
monopoly elements. Both long-distance services (contestable)and equipment manufacture and supply (competitive) were largely deregulated, while local telephone exchanges were deemed to be regional natural
Similar problems with
sectoral definitions have not been well recognized in other contexts,however. For
example, electricity generation and transmission have typically been treated as part
o f an electrical utility natural monopoly, although the evidence suggests that in many
circumstances only transmission has the required cost and demand characteristics to
be considered a natural monopoly.27(Running multiple sets o f transmission lines
across the countryside would generally be inefficient; having multiple firms generating electricity would not be.)
Another d~mension,apart from the sectoral, where the problem o f defining
natural monopoly arises is the spatial. Over what spatial area does the natural monopoly exist? Almost all natural monopoly regulation corresponds to existing city,
county, state, and federal boundaries; but the economic reality o f a natural rnonopoly knows no such bounds-it is purely an empirical question how far (spatially) the
natural monopoly extends. Again, the appropriate spatial boundaries o f a natural monopoly can change with changes in technology.

X-Inefficiency Resulting from Limited Competition

Thus far, we have described the potential social costs o f natural monopoly,
whether it prices to either maximize profits or cover cost, in terms o f deadweight
loss caused by allocational inefficiency. The social costs, however, may be larger be-

26~enneth
Roblnson, "Maxlm~zlngthe Publlc Benefits o f the AT&T Breakup " Journal of Pol~cy
Analysrs and Management, Vol 5 no 3, 1986. pp 572-97 For d~scuss~on
o f technolog~calchange that
eroded the natural monopoly characterlstlcs of telephone services, see I r w n Manley. Telecommunrcatrons

Figure 5.10 Shifting Demand and Multiple Firm Survival (Source:

William G. Shepherd, The Economics of Industrial Organization
(Englewood Cliffs, N.J.: Prentice Hall, 1979), fig. 3.4, p. 59.)

Amenca Markets Wrthout Boundaries (Westport. Conn Quorum, 1984)

27 For the case agalnst treatlng any aspects of the electrlc Industy as a natural monopoly, see
Robert W Poole. Jr , Unnatural Monopolres The Case for Deregulatrng Publrc Utrlrtles (Lexlngton. Mass
Lex~ngtonBooks, 1985)

Rationales for Public Policy: Market Failures

Chap. 5

'-

a,
.U

MC,,

Social surplus loss from efficient monopoly:

Minimum social surplus loss from inefficient monopoly:
Maximum social surplus loss from inefficient monopoly:
Figure 5.1 1

abc
aeg
net

+MC,fgPo

Information Asymmetry

107

observed from the behavior o f the firm. The actual deadweight loss, therefore, is at
least equal to the larger triangular aeg-the additional loss o f cbeg results because
.,
output is Qmxrather than Q
Some or all o f the very lightly shaded area MC,fgPo represents unneceSsary
cost that should be counted as either producer surplus or deadweight 1oss.i I f marginal costs are higher than the minimum because the managers o f the firms Gmploy
more real resources such as hiring workers who stand idle, then the area MC,fgPo
should be thought o f as a social surplus loss. If, on the other hand, costs are higher
because the managers pay themselves and their workers higher than necessary
wages, we should consider this area as rent-it represents unneessary payments
rather than the misuse o f real resources.,lf, however, potential employees and managers spend time or other resources attempting to secure the rent (say, by enduring
periods o f unemployment while waiting for one o f the overpaid jobs to open up),
then the rent may be dissipated and thus converted to deadweight loss.
Note that in the case o f an unregulated natural monopoly, one would not expect X-inefficiency to occur unless there were a divergence o f information and interests between owners o f the firm and its managers. We put this type o f problem into
the perspective of agency theory in our subsequent discussion o f government failures.
In summary, natural monopoly inherently involves the problem o f undersupply
by the market. The extent o f undersupply depends on the particular cost-anddemand conditions facing the monopolist, and the extent to which the market can
be contested. Natural monopoly may involve additional social surplus losses because the absence o f competition permits production at greater than minimum cost
to persist.

X-Inefficiency under Natural Monopoly

INFORMATION ASYMMETRY

! cause natural monopolies do not face as strong incentives as competitive firms to op-

\
i

erate at minimum cost. One o f the greatest advantages o f a competitive market is

that it forces firms to keep their costs down-in other words, the whole average and
marginal cost curves are as low as they can possibly be. In the absence o f competition, firms may be able to survive without operating at minimum cost. Harvey
Leibenstein coined the phrase X-inefficiency to describe the situation in which a mo(As we
nopoly does not achieve the minimum costs that are technically feasib~e.~'
shall see, X-inefficiency is not fully descriptive because transfers as well as technical
inefficienc~es
are often invois-ed.)
In Figure 5. i 1 we incorporate X-inefficiency into the basic analysis o f monopoly pricing. But first suppose that we ignore X-inefficiency for a moment. The deadweight loss associated with a profit-maximizing natural monopoly is the darkly
shaded triangular area abc (already discussed in Figure 5.9). If we take X-inefficiency
into account, however, the minimum possible marginal cost curve is lower than that

"See Harvey J. Leibenstem. Beyond Economic Man (Cambridge, MA: Harvard University Press.
1976) and Roger S. Frantz. X-Eficrency: Theov. Evidence and Appl~cations(Boston: Kluwer Academic
Publishers, 1988).

Please note that we do not use the title "information costs" or "imperfect information" in this section. The reason is that information is involved in market failure in at
least two distinct ways. First, information itself has public good characteristics. Consumption o f information is nonrivalrousdne person's consumption does not interfere with another's; the relevant analytical questlon is primarily whether exclusion is
or is not possible. Thus, in the public goods context, we are interested in the production and consumption o f information itself. Second--and the subject o f our discussion here-there may be situations where the amount o f lnformation about the
characterist~cs
o f a good varies in relevant ways across persons. The buyer and the
seller in a market transaction, for example, mav have different information about
the quality o f the good being traded. Similarly, there may be differences in the level
o f information relating to the attributes o f an externality between the generator o f
the externality and the affected party--workers, for instance, may not be as well informed about the health risks o f industrial chemicals as their employers. Notice that
in this context, we are not primarily interested in information as a good, but in the
degree o f asymmetry in the information relevant parties have about any good. We
thus distinguish between information as a good (the public good case) and information about a good's attributes as distributed between buyer and seller or between externality generator and affected party (the information asymmetry case).

Rationales for Public Policy: Market Failures

Social surplus loss from efficient monopoly:

Minimum social surplus loss from inefficient monopoly:
Maximum social surplus loss from inefficient monopoly:

Chap. 5

abc
aeg
ae&+ MC,agPO

Information Asymmetry

107

observed from the behavior o f the firm. T h e actual deadweight loss, therefore, is at
least equal to the larger triangular aeg-the
additional loss o f cbeg results because
output is Q, rather than Q,.
Some or all of the very lightly shaded area MC,,fgPo represents unnecessary
cost that should be counted as either producer surplus or deadweight loss.l If marginal costs are higher than the minimum because the managers o f the firms employ
more real resources such as hiring workers w h o stand idle, then the area MC,fgPo
should be thought o f as a social surplus loss. If, on the other hand, costs are higher
because the managers pay themselves and their workers higher than necessary
wages, w e should consider this area as rent-it
represents unnecessary payments
rather than the misuse o f real resources.flf,.however, potential employees and managers spend time or other resources attempting to secure the rent (say, by enduring
periods o f unemployment while waiting for one o f the overpaid jobs t o open up),
then the rent may be dissipated and thus converted t o deadweight loss.
Note that in the case o f an unregulated natural monopoly, one would not expect X-inefficiency t o occur unless there were a divergence o f information and interests between owners o f the firm and its managers. W e put this type o f problem into
the perspective o f agency theory in our subsequent discussion o f government failures.
In summary, natural monopoly inherently involves the problem o f undersupply
by the market. The extent o f undersupply depends on the particular cost-anddemand conditions facing the monopolist, and the extent t o which the market can
be contested. Natural monopoly may involve additional social surplus losses because the absence o f competition permits production at greater than minimum cost
to persist.

Figure 5.1 1 X-Inefficiency under Natural Monopoly

INFORMATION ASYMMETRY

cause natural monopolies do not face as strong incentives as competitive firms to operate at minimum cost. One o f the greatest advantages o f a competitive market is
that i t forces firms t o keep their costs down-in other words, the whole average and
marginal cost curves are as low as they can possibly be. In the absence o f competition, firms may be able to survive without operating at minimum cost. Harvey
Leibenstein coined the phrase X-ineficiency t o describe the situation in which a monopoly does not achieve the minimum costs that are technically feasible.28(As w e
shall see, X-inefficiency is not fully descriptive because transfers as well as technical
inefficiencies are often invoived.)
In Figure 5.11 w e incorporate X-ineficiency into the basic analysis o f monopoly pricing. But first suppose that w e ignore X-inefficiency for a moment. The deadweight loss associated with a profit-maximizing natural monopoly is the darkly
shaded triangular area abc (already discussed in Figure 5.9). l f w e take X-inefficiency
into account, however, the minimum possible marginal cost curve is lower than that

"See Harvey J. Leibenstein.Beyond Economic M a n (Cambridge, M A : Harvard University Press.

1976) and Roger S. Frantz. X-Eficiency: Theory. Ewjdence and Applications (Boston: Kluwer Academic
Publishers. 1988).

Please note that w e do not use the title "information costs" or "imperfect information" in this section. The reason is that information is involved in market failure in at
least t w o distinct ways. First, information itself has p&lic good characteristics. Conperson's consumption does not intersumption o f information is nonrivalrous-ne
fere w i t h another's; the relevant analytical question is primarily whether exclusion is
or is not possible. Thus, in the public goods context, w e are interested in the production and consumption o f information itself. Second-and the subject o f our discussion here-there may be situations where the amount o f information about the
characteristics o f a good varies in relevant ways across persons. T h e buyer and the
seller in a market transaction, for example, may have different information about
the quality o f the good being traded. Similarly, there may be differences in the level
of information relating t o the attributes o f an externality between the generator o f
the externality and the affected party-workers, for instance, may not be as well informed about the health risks o f industrial chemicals as their employers. Notice that
in this context, w e are not primarily interested in information as a good, but in the
.degree o f asymmetry in the information relevant parties have about any good. W e
thus distinguish between information as a good (the public good case) and information about a good's attributes as distributed between buyer and seller or between externality generator and affected party (the information asymmetry case).

Rationales for Public Policy: Market Failures

Uninformed consumption:
Informed consumption:
Deadweight loss if uninformed:
Extra producer rent if uninformed:
Figure 5.12

Chap. 5

QU
QI
nbc
PubnPl

Consumer Surplus Loss from Uninformed Demand

Inefficiency Due to lnformation Asymmetry

Figure 5.12 illustrates the potential social surplus loss associated with information asymmetry.29DUrepresents the quantities o f some good that a consumer would
purchase at various prices in the absence o f perfect information about its quality. I t
can therefore be thought o f as the consumer's uninformed demand schedule. Dl represents the consumer's informed demand schedule-the amounts o f the good that
would be purchased at various prices if the consumer were perfectly informed about
its quality. The quantity actually purchased by the uninformed consumer is deter-

'OThe baslc analysls was introduced by Sam Peltzman. "An Evaluation o f Consumer Protection
Legislation: The I962 Drug Amendments," Journal of Pol~t~cal
Economy. Vol. 81. no. 5, 1973. pp.
1049-91. For a discussion o f the empirical problems one encounters in using this approach when some
consumers overestimate and others underestimate the quality o f some good, sc ; Thomas McGuire.
R~chardNelson, and Thomas Spavins. "'An Evaluation o f Consumer Protection Legislation: The 1962
Drug Amendments': A Comment." JournalofPol~t~cal
Economy, Vol. 83, no. 3. 1975. pp. 655-61.

Information Asymmetry

109

mined by the intersection of Du with the supply schedule, 5.This amount, QU, is
greater than Q,, the amount that the consumer would have purchased if fully informed about the quality o f the good. The darkly shaded area abc equals the deadweight loss in consumer surplus resulting from the overconsumption. (For each unit
purchased beyond QI, the consumer pays more than its marginal value as measured
by the height o f the informed demand schedule.) This excess consumption also results in a higher equilibrium price (PU) which transfers surplus equal to the area
PubaP, from the consumer to the producer o f the good. Figure 5.12 signals the presence o f information asymmetry if the producer could have informed the consumer
about the true quality o f the good at a cost less than the deadweight loss in consumer surplus resulting when the consumer remains uninformed. More generally,
we have market failure due to information asymmetry when the producer does not
supply the amount o f information that maximizes the difference between the reduction in deadweight loss and the cost o f providing the information.
The same sort o f reasoning would apply if the consumer underestimated
rather than overestimated the quality o f the good. The consumer would suffer a
deadweight loss resulting from consuming less than QI. The incentives that the producer faces to provide the information, however, can be quite different in the two
cases. In the case where the consumer overestimates quality, providing information
results in a lower price and therefore a smaller transfer o f surplus from the consumer
to the producer-an apparent disincentive to provide the information. In the case
where the consumer underestimates quality, providing information results in a higher
price that increases producer surplus-the prospect o f this gain may encourage the
producer to supply information. As we discuss later, however, this incentive to provide information can be muted i f producers are unable to get consumers to distinguish their products from those o f competitors.

Diagnosing lnformation Asymmetry

Our first task in deciding when information asymmetry is likely to lead to market failure is to classify goods into useful categories. Economists have generally divided goods into two categories: search goods and experience goods.30A good is a
search good i f consumers can determine its characteristics with certainty prior to
purchase. For example, a chair is a search good because consumers can judge its
quality through inspection prior to purchase. A good is an experience good i f consumers can determine its characteristics only after purchase; examples include
meals, hair styling, concerts, legal services, and used automobiles. W e add a third
category, which we call post-experience goods, to distinguish those goods for which it
is difficult for consumers to determine quality even after they have begun consumption. For example, people may fail to associate adverse health effects with drugs that
they are consuming. Experience goods and post-experience goods differ primarily in
terms o f how effectively consumers can learn about quality through consumption.
After some period o f consumption, the quality o f the experience goods generally be-

'@The dlsttnction between search and experience goods was introduced by Philip Nelson, "lnformation and Consumer Behavior," Journal of Polit~calEconomy, Vol. 78, no. 2, 1970, pp. 311-29.

,%

,,

'

Rationales for Public Policy: Market Failures

Uninformed consumption:
Informed consumption:
Deadweight loss if uninformed:
Extra producer rent if uninformed:
Figure 5.12

Chap. 5

QU
QI
nbc
PubnPI

Consumer Surplus Loss from Uninformed D e m a n d

Inefficiency Due to lnformation Asymmetry

F~gure5.12 illustrates the potential social surplus loss associated w i t h information asymmetry.29DUrepresents the quantities o f some good that a consumer would
purchase at various prices In the absence o f perfect information about ~ t squality. It
can therefore be thought o f as the consumer's uninformed demand schedule. D, represents the consumer's informed demand schedule-the amounts o f the good that
would be purchased at various prices i f the consumer were perfectly informed about
its quality. The quantity actually purchased by the uninformed consumer is deter-

'"The bastc analysts was Introduced by Sam Peltzman, "An Evaluation of Consumer Protectton
Legtslatton The 1962 Drug Amendments." Journal of Politrcal Economy. Vol 81, no 5, 1973, pp
1049-91 For a dtscuss~ono f the emplrtcal problems one encounters In ustng thls approach when some
consumers overesttmate and others underestimate the qualtty o f some good, sr . Thomas McGutre.
R~chardNelson, and Thomas Spavtns. "'An Evaluatton o f Consumer Protect~onLegtslat~onThe 1962
Drug Amendments' A Comment " Journal ofPolrtrcal Economy. Vol 83, no 3, 1975, pp 655-61

Information Asymmetry

109

mined by the intersection o f DUwith the supply schedule, S.This amount. QU, is
greater than Q1, the amount that the consumer would have purchased if fully informed about the quality o f the good. The darkly shaded area abc equals the deadweight loss in consumer surplus resulting from the overconsumption. (For each unit
purchased beyond QI, the consumer pays more than its marginal value as measured
by the height o f the informed demand schedule.) This excess consumption also results in a higher equilibrium price (PU) which transfers surplus equal t o the area
PUbaP, from the consumer t o the producer o f the good. Figure 5.12 signals the presence o f information asymmetry if the producer could have informed the consumer
about the true quality o f the good at a cost less than the deadweight loss in consumer surplus resulting when the consumer remains uninformed. More generally,
w e have market failure due t o information asymmetry when the producer does not
supply the amount o f inforrnat~onthat maximizes the difference between the reduction in deadweight loss and the cost o f providing the information.
The same sort o f reasoning would apply if the consumer underestimated
rather than overestimated the quality o f the good. The consumer would suffer a
deadweight loss resulting from consuming less than QI. The incentives that the producer faces t o provide the information, however, can be quite different in the t w o
cases. In the case where the consumer overestimates quality, providing information
results in a lower price and therefore a smaller transfer o f surplus from the consumer
t o the producer-an apparent disincentive t o provide the information. In the case
where the consumer underestimates quality, providing information results in a higher
price that increases producer surplus-the prospect o f this gain may encourage the
producer t o supply information. A s w e discuss later, however, this incentive t o provide information can be muted i f producers are unable t o get consumers t o distinguish their products from those o f competitors.

Diagnosing lnformation Asymmetry

O u r first task in deciding when information asymmetry is likely t o lead t o market failure is t o classifj, goods into useful categories. Economists have generally divided goods into t w o categories: search goods and experrence goods.30 A good is a ,,
search good if consumers can determine its characteristics with certainty prior t o ,
purchase. For example, a chair is a search good because consumers can judge its 1
quality through inspection prior t o purchase. A good is an experience good if con- .'
sumers can determine its characteristics only after purchase; examples include
meals, hair styling, concerts, legal services, and used automobiles. W e add a third
category, which w e call post-experience goods, t o distinguish those goods for which it
is difficult for consumers t o determine quality even after they have begun consumption. For example, people may fail t o associate adverse health effects with drugs that
they are consuming. Experience goods and post-experience goods differ primarily in
terms o f h o w effectively consumers can learn about quality through consumption.
After some period o f consumption, the quality o f the experience goods generally be-

'%e
dlsttnction between search and experience goods was Introduced by Philtp Nelson, "lnformatton and Consumer Behavior." JournalofPolitrcol Economy, Vol. 78, no. 2. 1970, pp. 311-29.

Rationales for Public Policy: Market Failures

.

Chap. 5

comes apparent; in contrast, continued consumption does not necessarily reveal t o

consumers the quality o f the post-experience good.3'
Within these three categories, a number o f other factors help determine
whether information asymmetry is likely t o lead t o serious market failure. The effectiveness o f any information gathering strategy, other things equal, generally depends
on the varrance in the quality o f units o f a good (heterogeneity) and thefrequency with
which consumers make purchases. The potential costs o f the information asymmetry t o consumers depends on the extent t o which they perceive the full price o f the
good, including imputed costs o f harm from use.32The cost of searching for candidate
purchases and the full price determine h o w expensive and potentially beneficial it is
for consumers t o gather information.

Search Goods. Searching can be thought o f as a sampling process in which

consumers incur costs to inspect units o f a good. A consumer pays a cost, Cs, to see
a particular combination o f price and quality. If the price exceeds the consumer's
marginal value for the good, no purchase is made and the consumer either again pays
CS t o see another combination o f price and quality or stops sampling. If the consumer's marginal valuation exceeds price, then the consumer either makes a purchase or pays Cs again in expectation o f finding a more favorable good in terms o f
the excess o f marginal value over price. When C, is zero, the consumer will find i t
advantageous to take a large sample and discover the complete distribution o f available price and quality combinations so that the pre-search information asymmetry
disappears. For larger C, however, the consumer will take smaller samples, other
things equal, so that information asymmetry may remain. Additionalty, because the
range in price for a given quality is likely to be positively correlated w i t h price, optimal sample sizes will be smaller the larger the ratio o f Cs t o expected price.
The more heterogeneous the available combinations o f price and quality, the
more likely that the consumer will fail t o discover a more favorable choice for any
given sample size. In contrast, even small samples will eliminate information asymmetry i f the price and quality combinations are highly homogeneous,..Once consumers realize that nearly identical units are offered at the same price, the optimal
sample size falls to one.
Going beyond a static view, the frequency o f purchase becomes important in
determining whether information asymmetry remains. Ifthe frequency o f purchase is
high relative t o the rate at which the underlying distribution o f combinations o f price
and quality change, then consumers accumulate information over time that reduces
the magnitude o f the information asymmetry. If the frequency o f purchase is low relative to the rate o f change in the underlying distribution, then accumulated informa-

Information Asymmetry

111

tion will not necessarily lead t o reductions in information asymmetry. In either case,
however, frequent purchasers may become more experienced searchers so that Cs
falls and larger samples become efficient.
Thus, i f search costs are small relative t o the expected purchase price,or the
distribution o f price and quality combinations is fairly homogeneous or the frequency
o f purchase is high relative t o the rate o f change in the distribution o f price and quali t y combinations, then in for ma ti^^ asymmetry is unlikely to lead t o significant ineficiency. In the case where search costs are high relative to the expected purchase
price, the distribution o f price and quality combinations is very heterogeneous, and
the frequency o f purchase is relatively low, information asymmetry may lead to significant inefficiency. But, i f it is possible for producers t o distinguish their products by
brand, they have an incentive t o undertake informative advertising that reduces
search costs for consumers. When brands are difficult to establish, as in the case of,
say, farm produce, retailers may act as agents for consumers by advertising prices
and qualities. Because the veracity o f such advertising can be readily determined by
consumers through inspection o f the goods, and because retailers and firms offering
brand name products have an incentive t o maintain favorable reputations, w e expect
the advertising generally to convey accurate and useful information.
In summary, -search goods rarely involve information asymmetry that leads t o
significant ahd-persistent inefficiency. When inefficiency does occur, it takes the
form o f consumers' forgoing purchases that they would have preferred t o ones that
they nevertheless found beneficial. From the perspective o f public policy, intervention in markets for search goods can rarely be justified on efficiency grounds.

"See Aidan R. Vining and David L. Weimer. "lnformation Asymmetry Favoring Sellers: A Policy
Framework." Policy Sciences. Vol. 21, no. 4, 1988, pp. 281-303.

Experience Goods: Primary Markets. Consumers can determine the

quality o f experience goods with certainty only through consumption. T o sample,
they must bear the search costs, Cs, and the full price, P* (the purchase price plus
the expected loss o f failure or damage collateral) with consumption.33The full price
o f consuming a meal at an unfamiliar restaurant, for instance, is the sum o f purchase
price (determined from the menu) and the expected cost o f any adverse health effects (ranging from indigestion t o poisoning) from the meal being defective. Of
course, even when the expected collateral loss is zero, prior to consumption, the
marginal value that the consumer places on the meal is not known w i t h certainty.
In contrast to search goods, where, holding search costs constant, consumers
optimally take larger samples for more expensive goods, they optrmally take smaller
samples for more expensive experience goods. Indeed, for all but the very inexpensive experience goods, we expect sampling (equivalent to the frequency o f purchase
for experience goods) t o be governed primarily by durability. For example, sampling
to find desirable restaurants will generally be more frequent than sampling t o find
good used automobiles.
As is the case with search goods, the more heterogeneous the quality o f an
experience good, the greater is the potential for inefficiency due to information

3 2 fomal
~
specificat~ono f the concept o f full price is provided by Walter Oi, "The Economics of
Product Safety," Bell Journal of Econom~csand Management Science. Vol. 4. no. 1. 1973, pp. 3-28. H e
considers the situation in which a consumer buys X units o f a good at price, P, per unit. I f the probability
that any one unit is defective is I - 9, then on average the consumer expects Z = qX good units. If each
bad unit infl~ctson average damage equal to W, then the expected total cost o f purchase is C = PX +
W(X - Z),
wh~chimpl~esa full price per good unit o f P* = C/Z = P/q + W(1 - q)/q.

''In our discussion o f informat~onasymmetry, we assume that consumers cannot sue producers
for damages (in other words. they do not enjoy a property right to safety). As we discuss ~nChapter 9
under "framework regulation." tort and contract law often reduce the inefficiency o f information asymmetry by deterring parties from withholding relevant information in market transactions.

.'

Rationales for Public Policy: Market Failures

Chap. 5

comes apparent; in contrast, continued consumption does not necessarily reveal to

consumers the quality of the post-experience good.31
Within these three categories, a number of other factors help determine
whether information asymmetry is likely t o lead t o serious market failure. T h e effectiveness of any information gathering strategy, other things equal, generally depends
on the variance in the quality of units of a good (heterogeneity) and thefrequency with
which consumers make purchases. T h e potential costs of the information asyrnmetry t o consumers depends on the extent t o which they perceive the full price of the
good, including imputed costs of harm from use.32T h e cost of searching for candidate
purchases and the full price determine how expensive and potentially beneficial it is
for consumers to gather information.
Search Goods. Searching can be thought of as a sampling process in which
consumers incur costs t o inspect units of a good. A consumer pays a cost, CS.to see
a particular combination of price and quality. If the price exceeds the consumer's
marginal value for the good, no purchase is made and the consumer either again pays
C5t o see another combination of price and quality or stops sampling. If the consumer's marganal valuation exceeds price, then the consumer either makes a purchase or pays C g again in expectation of finding a more favorable good in terms of
the excess of marginal value over price. When C5is zero, the consumer will find it
advantageous t o take a large sample and discover the complete distribution of available price and quality combinations so that the pre-search information asymmetry
disappears. For larger C,,however, the consumer will take smaller samples, other
things equal, so that information asymmetry may remain. Additionalfy. because the
range in price for a given quality is likely to be positively correlated with price, optimal sample sizes will be smaller the larger the ratio of Cst o expected price.
T h e more heterogeneous the available combinations of price and quality, the
more likely that the consumer will fail t o discover a more favorable choice for any
given sample size. In contrast, even small samples will eliminate information asymmetry if the prlce and quality combinations are highly homogeneous. Once consumers realize that nearly identical units are offered a t the same price, the optimal
sample size falls to one.
Going beyond a static view, the frequency of purchase becomes important in
determining whether information asymmetry remains. If the frequency of purchase is
high relative t o the rate a t which the underlying distribution of combinations of price
and quality change, then consumers accumulate information over time that reduces
the magnitude of the information asymmetry. If the frequency of purchase is low relative to the rate of change in the underlying distribution, then accumulated informa-

Information Asymmetry

11 1

tion will not necessarily lead t o reductions in information asymmetry. In either case,
however, frequent purchasers may become more experienced searchers so that C5
falls and larger samples become efficient.
Thus, if search costs are small relative t o the expected purchase price,or the
distribution of price and quality combinations is fairly homogeneous or thefrequency
of purchase is high relative t o the rate of change in the distribution of price and quality combinations, then informatio; asymmetry is unlikely to lead t o significant inefficiency. In the case where search costs are high relative t o the expected purchase
price, the distribution of price and quality combinations is very heterogeneous, and
the frequency of purchase is relatively low, information asymmetry may lead to significant inefficiency. But, if it is possible for producers to distinguish their products by
brand, they have an incentive t o undertake informative advertising that reduces
search costs for consumers. When brands are difficult t o establish, as in the case of,
say, farm produce, retailers may act as agents for consumers by advertising prices
and qualities. Because the veracity of such advertising can be readily determined by
consumers through inspection of the goods, and because retailers and firms offering
brand name products have an incentive t o maintain favorable reputations, we expect
the advertising generally t o convey accurate and useful information.
In summary,-search goods rarely involve information asymmetry that leads t o
significant ahd. persistent inefficiency. When inefficiency does occur, it takes the
form of consumers' forgoing purchases that they would have preferred to ones that
they nevertheless found beneficial. From the perspective of public policy, intervention in markets for search goods can rarely be justified on efficiency grounds.

"See Aidan R. Vining and David L. We~mer."lnformation Asymmetry Favoring Sellers: A Policy
Framework." Poiicy Sciences, Vol. 21, no. 4. 1988, pp. 281-303.

Experience Goods: Primary Markets. Consumers can determine the

quality of experience goods with certainty only through consumption. T o sample,
they must bear the search costs, C5 and the full price, P* (the purchase price plus
the expected loss of failure or damage collateral) with consumption.33 he full price
of consuming a meal a t an unfamiliar restaurant, for instance, is the sum of purchase
price (determined from the menu) and the expected cost of any adverse health effects (ranging from indigestion to poisoning) from the meal being defective. O f
course, even when the expected collateral loss is zero, prior t o consumption, the
marginal value that the consumer places on the meal is not known with certainty.
In contrast t o search goods, where, holding search costs constant, consumers
optimally take larger samples for more expensive goods, they optimally take smaller
samples for more expensive experience goods. Indeed, for all but the very Inexpensive experience goods, w e expect sampling (equivalent t o the frequency of purchase
for experience goods) to be governed primarily by durability. For example, sampling
to find desirable restaurants will generally be more frequent than sampling to find
good used automobiles.
A s is the case with search goods, the more heterogeneous the quality of an
experience good, the greater is the potential for inefficiency due t o information

''A formal specification o f the concept o f full price is provided by Walter Oi. "The Economics of
Product Safety." Bell Journal of Economics and Management Science. Vol. 4, no. 1. 1973, pp. 3-28. H e
considers the situation in which a consumer buys X units o f a good at price, P, per unlt. If the probabil~ty
that any one unit 1s defective is I - q, then on average the consumer expects Z = qX good units. If each
bad unit inflicts on average darna~eequal to W, then the expected total cost o f purchase is C = P X +
W(X - Z),
which implies a full prlce per good unit o f P* = C/Z = P/q + W(I - q)/q.

331nour discussion o f information asymmetry, we assume that consumers cannot sue producers
for damages (in other words, they do not enjoy a property right to safety). As we discuss In Chapter 9
under "framework regulation." tort and contract law often reduce the inefficiency o f ~nformationasymmetry by deterring parties from withholding relevant information in market transactions.

Rationales for Public Policy: Market Failures

Chap. 5

asymmetry. T h e consumption of an experience good, however, may involve more

than simply forgoing a more favorable purchase of the good. Once consumption reveals quality, the consumer may discover that the good provides less marginal value
than its price and therefore regret having made the purchase regardless of the availability of alternative units of the good. The realized marginal value may actually be
negative ~fconsumption causes harm.
Learning from the consumption of experience goods varies in effectiveness. If
the quality of the good is homogeneous and stable, then learning is complete after
the first consumption--consumers know how much marginal value they will derive
from their next purchase including the expected loss from product failure or collateral damage. If the quality is heterogeneous or unstable, then learning proceeds more
slowly. Unless consumers can segment the good into more homogeneous groupings,
say by brands or reputations of sellers, learning may result only in better ex ante estimates of the mean and variance of their ex post marginal valuations. When consumers can segment the good into stable and homogeneous groupings, repeated
sampling helps them discover their most preferred sources of the good in terms of
the memiand variance of quality. For example, national motel chains with reputations for standardized management practices may provide travelers with a low variance alternative t o independent motels in unfamiliar locales.
In what situations can informative advertising play an important role in reducing information asymmetry? Generally, informative advertising can be effective
when consumers correctly believe that sellers have a stake in maintaining reputations for providing reliable information. A seller who invests heavily in developing a
brand name with a favorable reputation is more likely t o provide =curate and useful
information than an unknown firm selling a new product or an individual owner selling a house or used automobile.
When consumers perceive that sellers do not have a stake in maintaining a
good reputation, and marginal cost of supply rises with quality, then a "lemons"
problem may arise: Consumers perceive a full price based on average quality so that
only sellers of lower than average quality goods can make a profit and survive in the
market.34 In the extreme, producers offer only goods of low quality.35
When reliability is an important element of quality, firms may offer warranties
that promise t o compensate consumers for a portion of replacement costs, collateral
darnage, or both. The warranties thus provide consumers with insurance against low
quality: Higher purchase prices include an implicit insurance premium and the potential loss from product failure is reduced. The quality of a warranty may itself be uncertain, however, if consumers do not know how readily firms honor their promises.
Further, firms may find it impractical t o offer extensive warranties in situations
where it is costly t o monitor the care taken by consumers. Nevertheless, warranties

I
I

34The"lemons" argument originated with George Akerlof. "The Market for Lemons." Quarterly
84, no. 3, 1970, pp. 488-500.
35Economjstshave recently considered the role of expenditures to establish reputation as a means
for high quality producers to distinguish themselves. See, for instance. Benjamin Klein and Keith B. Leffler.
"The Role of Market Forces in Assuring Contractural Performance."Journal ofPol;tical Economy. Vol.
89, no. 4, 1981, pp. 61541: and Paul Milgrom and John Roberts. "Price and Advertising Signals of Product Quality." JournolofPoliticol Economy. Vol. 94, no. 4. 1986. pp. 79621

Journalof Economics, Vol.

Information Asymmetry

113

serve as a common device for reducing the consequences of information asymmetry

for experience goods.

Experience Goods: T h e Role of Secondary Markets. Producers and

consumers often turn t o private third parties t o help remedy information asymmetry
problems. Certification services, agents, subscription services, and loss control by insurers are the most common market responses that arise.
Certification services "guarantee" minimum quality standards in processes or
products. Professional associations, for instance, often set minimum standards of
training or experience for their members-the board-certified specialities in medicine
are examples. Perhaps closer t o most people's experience, the Better Business Bureau requires members to adhere t o a code of fair business practices. Underwriters
Laboratories tests products against minimum fire safety standards before giving its
seal of approval.36When such services establish their own credibility, they help producers t o distinguish their goods satisfying the minimum standards from goods that
do not.
Agents often sell advice about the qualities of expensive, infrequently purchased, heterogeneous goods. These agents combine expertise with learning from
being participants in a large number of transactions between different pairs of buyers and sellers. For example, most people d o not frequently purchase houses,
whlch are expensive and heterogenous in quallty. Because owners typically enjoy
an informational advantage by virtue of their experiences living in their houses,
prospective buyers often turn t o engineers and architects t o help assess structural
integrity. Art and antique dealers, jewelers, and general contractors provide similar
services.
T h e problem of excluding nonpaying users limits the supply of agents for goods
that are homogeneous-ne
consumer can easily pass along information that the
agent provides t o prospective purchasers of similar units of the good. Inexpensive
goods are unlikely to provide an adequate incentive for consumers t o pay for the advice of agents. (Note that the full price is the relevant r n e a s u r e q n e might very well
be willing to pay for a visit t o a doctor t o get advice about a drug with a low purchase price but a high potential for harm t o one's health.) Finally, agents are less
likely t o be relatively attractive for goods with high frequencies of purchase because
consumers can ofien learn effectively on their own.
Consumers often rely on the experiences of friends and relatives t o gather information about the quality of branded products. They may also be willing to pay for
published informat~onabout such products. But such subscription services, which
have public good characterist~cs,are likely t o be undersupplied from the perspective
of economic efficiency because nonsubscribers can often free ride on subscribers by
interrogating them and by borrowing their magazines. (The very existence of subscription services such as Consumer Reports suggests that a large number of consumers see the marginal costs of mooching or using a library as higher than the subscription price.)

36Foran overview of private organizations that set quality standards, see Ross E. Cheit, Setting
Safity Standords: Regulatron in the Public and Private Sectors (Berkeley: University of California Press,
1990).

Rationales for Public Policy: Market Failures

Chap. 5

asymmetry. T h e consumption of an experience good, however, may involve more

than simply forgoing a more favorable purchase of the good. Once consumption reveals quality, the consumer may discover that the good provides less marginal value
than its price and therefore regret having made the purchase regardless of the availability of alternative units of the good. T h e realized marginal value may actually be
negative if consumption causes harm.
Learning from the consumption of experience goods varies in effectiveness. If
the quality of the good is homogeneous and stable, then learning is complete after
the first consumption-consumers know how much marginal value they will derive
from their next purchase including the expected loss from product failure or collateral damage. If the quality is heterogeneous or unstable, then learning proceeds more
slowly. Unless consumers can segment the good into more homogeneous groupings,
say by brands or reputations of sellers, learning may result only in better ex ante estimates of the mean and variance of their ex post marginal valuations. When consumers can segment the good into stable and homogeneous groupings, repeated
sampling helps them discover their most preferred sources of the goad in terms of
the mean,,and variance of quality. For example, national motel chains with reputations for s'tandardized management practices may provide travelers with a low variance alternative t o independent motels in unfamiliar locales.
In what srtuations can informative advertising play an important role in reducing information asymmetry? Generally, informative advertising can be effective
when consumers correctly believe that sellers have a stake in maintaining reputations for providing reliable information. A seller w h o invests heavily in developing a
brand name with a favorable reputation is more likely t o provide *curate and useful
information than an unknown firm selling a new product or an individual owner selling a house or used automobile.
When consumers perceive that sellers d o not have a stake in maintaining a
good reputation, and marginal cost of supply rises with quality, then a "lemons"
problem may arise: Consumers perceive a full price based on average quality so that
only sellers of lower than average quality goods can make a profit and survlve in the
market.34 In the extreme, producers offer only goods of low quality.35
When reliability is an important element of quality, firms may offer warranties
that promise t o compensate consumers for a portion of replacement costs, collateral
damage, or both. T h e warranties thus provide consumers with insurance against low
quality: Higher purchase prices include an impltcit insurance premium and the potential loss from product failure is reduced. T h e quality of a warranty may itself be uncertain, however, if consumers do not know how readily firms honor their promises.
Further, firms may find it impractical t o offer extensive warranties in situations
where it is costly t o monitor the care taken by consumers. Nevertheless, warranties

34The"lemons" argument originated with George Akerlof. "The Market for Lemons," Quarterly
Journal of Economjcs. Vol. 84, no. 3. 1970, pp. 488-500.

3SEconomistshave recently considered the role o f expenditures to establish reputation as a means

for high quality producers to distinguish themselves. See, for instance. Benjamin Klein and Keith B. Leffler.
"The Role o f Market Forces in Assuring Contractural Performance." Journal of Polttical Economy. Val.
89, no. 4. 1981, pp. 615-41; and Paul Mtlgrom and John Roberts. "Price and Advertising Signals o f Prcduct Quality," JournalofPolit~cal Economy. Vol. 94. no. 4. 1986, pp. 796-21.

Information Asymmetry

113

serve as a common device for reducing the consequences of information asymmetry

for experience goods.

Experience Goods: The Role of Secondary Markets. Producers and

consumers often turn t o private third parties t o help remedy information asymmetry
problems. Certification services, agents, subscription services, and loss control by insurers are the most common market responses that arise.
Certification services "guarantee" minimum quality standards in processes or
products. Professional associations, for instance, often set minimum standards of
training or experience for their members-the board-certified specialities in medicine
are examples. Perhaps closer t o most people's experience, the Better Business Bureau requires members t o adhere t o a code of fair business practices. Underwriters
Laboratories tests products against minimum fire safety standards before giving its
seal of approval.36When such services establish their own credibility, they help producers t o distinguish their goods satisfying the minimum standards from goods that
d o not.
Agents often sell advice about the qualities of expensive, infrequently purchased, heterogeneous goods. These agents combine expertise with learning from
being participants in a large number of transactions between different pairs of buyers and sellers. For example, most people d o not frequently purchase houses,
which are expensive and heterogenous in quality. Because owners typically enjoy
an informational advantage by virtue of their experiences living in their houses,
prospective buyers often turn t o engineers and architects t o help assess structural
integrity. Art and antique dealers, jewelers, and general contractors provide similar
services.
T h e problem of excluding nonpaying users limits the supply of agents for goods
that are homogeneous-ne
consumer can easily pass along information that the
agent provides t o prospective purchasers of similar units of the good. Inexpensive
goods are unlikely t o provide an adequate incentive for consumers t o pay for the advice of agents. (Note that the full price is the relevant measure-ne
might very well
be willing t o pay for a visit t o a doctor t o get advice about a drug with a low purchase price but a high potential for harm to one's health.) Finally, agents are less
likely t o be relatively attractive for goods with high frequencies of purchase because
consumers can often learn effectively on their own.
Consumers often rely on the experiences of friends and relatives t o gather information about the quality of branded products. They may also be willing t o pay for
published information about such products. But such subscription services, which
have public good characteristics, are likely t o be undersupplied from the perspective
of economic eficiency because nonsubscribers can often free ride on subscribers by
interrogating them and by borrowing their magazines. (The very existence of subscription services such as Consumer Reports suggests that a large number of consumers see the marginal costs of mooching or using a library as higher than the subscription price.)

j6For an overview of private organizations that set quality standards, see Ross E. Cheit, Setting
Safety Standards: Regulation in the Public ond Private Sectors (Berkeley: University o f California Press,
1990).

Rationales for Public Policy: Market Failures

Chap. 5

Insurers sometimes provide information to consumers as part o f their efforts t o

limit losses. For example, health maintenance organizations (HMOs), which provide
medical insurance, often publish newsletters that warn members o f potentially dangerous and unhealthy goods such as tanning salons and certain diet products. Casualty insurers may signal warnings through their premiums as well as through direct
information. Fire insurance underwriters, for instance, set premiums based on the
types of equipment businesses plan t o use; they also often inspect commercial properties t o warn proprietors o f dangerous equipment and inform them about additions
that could reduce their risks o f fire.

Experience Goods: Summary.

By their very nature, experience goods

offer the potential for serious inefficiency caused by information asymmetry. Secondary markets, however, limit the inefficiency for many experience goods by facilitating consumer learning and providing incentives for the revelation o f product qual~ t y Nevertheless,
.
problems are likely t o remain in t w o sets o f circumstances:,first,
when quality is highly heterogeneous, branding is ineffective, and agents are either
unavailable or expensive relative t o the full price o f the good; and second, where the
distribution of quality is unstable so that consumers and agents have difficulty learning effectively. In these fairly limited circumstances, market failure may justify public
intervention on efficiency grounds.
Consumption does not perfectly reveal the true
quality o f post-experience goods.37Quality remains uncertain because the individual
consumer has dificulty recognizing the causality between consufiption and some of
its effects. For example, consumers may not recognize that the fumes o f a cleaning
product increase their risks o f liver cancer because they do not expect the product t o
have such an effect. Often, the effects o f consumption only appear after great delay
so that consumers do not connect them with the product. An extreme example is
DES, a drug that increases the risk o f cancer in mature daughters o f women who
used it during pregnancy. Many other drugs require extended periods before all their
effects are manifested. Young smokers, for instance, may fully appreciate the addictive power o f tobacco only after they are addicted. Beyond drugs, many medical services appear t o be post-experience goods-patients often cannot clearly link the
state o f their health t o the treatments that they have received.
In terms o f our earlier discussion, the consumer o f a post-experience good may
not have sufficient knowledge o f the product to form reasonable estimates o f its full
price even after repeated consumption. In other words, whereas experience goods
involve risk (known contingencies with known probabilities o f occurrence), postexperience goods involve uncertainty (unknown cont~ngenciesor unknown probabilities o f occurrence). Further, repeated consumption o f a post-experience good does
not necessarily lead t o accurate estimates o f risk.

Post-Experience Goods.

"Our category of post-experience goods shares characteristics with Arrow's "trust goods" and
captures Darby and Karni's "credence qualit~es."Kenneth J. Arrow. "Uncertainty and the Welfare ECOn m i c s of Medical Care." American Economic Review. Vol. 53, no. 5, 1963, pp. 941-73; Michael R. Dahy
and Edi Karni. "Free Competition and the Optimal Amount of Fraud," Journal of Low ond Econmics.
Vol. 16, no. I . 1973,pp. 67-88.

Conclusion
Obviously, the potential for substantial and persistent inefficiency due t o information asymmetry is greater for post-experience than experience goods. Other
things equal, frequent purchases are less effective in eliminating information asymmetry for post-experience goods. In contrast t o the case o f experience goods, information asymmetry can persist for an extended period for even homogeneous postexperience goods.
Turning t o private responses t o the information asymmetry inherent in the primary markets for post-experience goods, we expect secondary markets in general to
be less effective than they are for experience goods because o f learning problems.
Nevertheless, secondary markets can play important roles. For example, consider
the services that are, or have been privately offered to provide information about
pharmaceuticals. The Medical Letter on Drugs and Therapeutics, one o f the few medical periodicals that covers its costs through subscriptions rather than advertising,
provides independent assessments o f all new drugs approved by the FDA. Many
hospitals, and some insurers, have formulary committees that review information
about drugs for participating physicians. Between 1929 and 1955, the American
Medical Association ran a seal-of-acceptance program that subjected all products
advertised in the Journd of the Amerlcan Medical Association to committee review.
The National Formulary and the United States Pharmacopoeiawere early attempts
at standardizing pharmaceutical products. Of course, most o f these information
sources are directed at physicians and pharmacists who commonly serve as agents
for consumers o f therapeutic drugs. Although their learning from direct observation
may not be effective, they can often determine i f advertising claims for products are
based on reasonable scientific evidence.

Summary of Information Asymmetry

The potential for inefficiency due to information asymmetry between buyers
and sellers can be found in many markets. The potential is rarely great for search
goods, ohen great for experience goods, and usually great for postexperience goods.
The extent to which the potential is actually realized, however, depends largely on
whether public goods problems hinder the operation o f secondary market mechanisms that provide corrective information. Thus, market failure is most likely in situations where information asymmetry in primary markets occurs in combination with
public good problems in secondary markets.

CONCLUSION
The traditional market failures involve circumstances in which the "invisible hand"
fails t o produce Pareto efficiency. They thus indicate possibilities for improving the
eficiency o f market outcomes through collective action. Along with the pursuit o f
distributional values, which we consider in Chapter 7, they constitute the most commonly advanced rationales for public policy. But other, less easily accommodated,
discrepancies between the competitive framework and the real world also suggest
opportunities for advancing efficiency through collective action. These discrepancies, the subject o f the next chapter, provide additional rationales for public policy.

Rationales for Public Policy: Market Failures

Chap. 5

Insurers sometimes provide information to consumers as part o f their efforts to

limit losses. For example, health maintenance organizations (HMOs), which provide
medical insurance, often publish newsletters that warn members o f potentially dangerous and unhealthy goods such as tanning salons and certain diet products. Casualty insurers may signal warnings through their premiums as well as through direct
information. Fire insurance underwriters, for instance, set premiums based on the
types o f equipment businesses plan t o use; they also often inspect commercial properties to warn proprietors o f dangerous equipment and inform them about add~tions
that could reduce their risks o f fire.

Experience Goods: Summary. By their very nature, experience goods

offer the potential for serious inefficiency caused by information asymmetry. Secondary markets, however, limit the inefficiency for many experience goods by facilitating consumer learning and providing incentives for the revelation o f product quality. Nevertheless, problems are likely to remain in t w o sets o f circumstances: first,
when quality is highly heterogeneous, branding is ineffective, and agents are either
unavailable or expensive relative t o the full price o f the good; and second, where the
distribution o f quality is unstable so that consumers and agents have difficulty learn;ng effectively. In these fairly limited circumstances, market failure may justi6 public
intervention on efficiency grounds.
Post-Experience Goods. Consumption does not perfectly reveal the true
quality o f post-experience goods.37Quality remains uncertain because the individual
consumer has difficulty recognizing the causality between c o n s u ~ p t i o n
and some o f
its effects. For example, consumers may not recognize that the fumes o f a cleaning
product increase their risks o f liver cancer because they do not expect the product to
have such an effect. Often, the effects o f consumption only appear after great delay
so that consumers do not connect them with the product. A n extreme example is
DES, a drug that increases the risk o f cancer in mature daughters o f women who
used i t during pregnancy. Many other drugs require extended periods before all their
effects are manifested. Young smokers, for instance, may fully appreciate the addictive power o f tobacco only after they are addicted. Beyond drugs, many medical services appear t o be post-experience goods-patients
often cannot clearly link the
state o f their health t o the treatments that they have received.
In terms of our earlier discussion, the consumer o f a post-experience good may
not have sufficient knowledge o f the product to form reasonable estimates o f its full
price even after repeated consumption. In other words, whereas experience goods
involve risk (known contingencies w ~ t hknown probabilities o f occurrence), postexperience goods involve uncertainty (unknown contingencies or unknown probabilities of occurrence). Further, repeated consumption o f a post-experience good does
not necessarily lead to accurate estimates o f risk.

370urcategory of post-experience gods shares characterlstlcs w ~ t hArrow's "trust goods" and

captures Darby and Karni's "credence qual~t~esKenneth J. Arrow, "Uncertainty and the Welfare Economics of Medlcal Care," American Economrc Review, Vol. 53, no. 5, 1963, pp 941-73; Michael R. Darby
and Edi Karni, "Free Competit~onand the Optimal Amount of Fraud." Journal of Law and Economics.
Vol. 16, no 1, 1973, pp. 67-88.
"

Conclusion

115

Obviously, the potential for substantial and persistent inefficiency due to information asymmetry is greater for post-experience than experience goods. Other
things equal, frequent purchases are less effective in eliminating information asymmetry for post-experience goods. In contrast t o the case o f experience goods, information asymmetry can persist for an extended period for even homogeneous postexperience goods.
Turning t o private responses t o the information asymmetry inherent in the primary markets for post-experience goods, w e expect secondary markets in general to
be less effective than they are for experience goods because o f learning problems.
Nevertheless, secondary markets can play important roles. For example, consider
the services that are, or have been privately offered to provide information about
pharmaceuticals. The Medical Letter o n Drugs a n d Therapeutics, one o f the few medical periodicals that covers its costs through subscriptions rather than advertising,
provides independent assessments o f all new drugs approved by the F D A . Many
hospitals, and some insurers, have formulary committees that review information
about drugs for participating physicians. Between 1929 and 1955, the American
Medical Association ran a seal-of-acceptance program that subjected all products
advertised in the Journal of the American Medical Association t o committee review.
The National Formulary and the United States Pharmacopoeia were early attempts
at standardizing pharmaceutical products. Of course, most o f these information
sources are directed at physicians and pharmacists who commonly serve as agents
for consumers o f therapeutic drugs. Although their learning from direct observation
may not be effective, they can often determine if advertising claims for products are
based on reasonable scientific evidence.

Summary of InformationAsymmetry
The potential for inefficiency due to information asymmetry between buyers
and sellers can be found in many markets. The potential is rarely great for search
goods, often great for experience goods, and usually great for post-experience goods.
The extent to which the potential is actually realized, however, depends largely on
whether public goods problems hinder the operation o f secondary market mechanisms that provide corrective informat~on.Thus, market failure is most likely in situations where information asymmetry in primary markets occurs in combination with
public good problems in secondary markets.

CONCLUSION
The traditional market failures involve circumstances in which the "invisible hand"
fails to produce Pareto efficiency. They thus indicate possibilities for improving the
efficiency o f market outcomes through collective action. Along with the pursuit o f
distributional values, which w e consider in Chapter 7, they constitute the most commonly advanced rationales for public policy. But other, less easily accommodated,
discrepancies between the competitive framework and the real world also suggest
opportunities for advancing efficiency through collective action. These discrepancies, the subject of the next chapter, provide additional rationales for public policy.