intermediate public economics 6 public goods - hsakamoto · payoff matrix • consider the payoff...
TRANSCRIPT
Intermediate public economics 6Public goods
Hiroaki Sakamoto
June 26, 2015
Contents
1. Definition and examples2. Modeling public goods
2.1 Model2.2 Efficient allocation and equilibrium
3. Lindahl mechanism3.1 Design3.2 Efficiency3.3 Practical relevance
4. VCG mechanism4.1 Design4.2 Demand-revealing property4.3 Inefficiency
Private goods, public goods
Private goods (as opposed to public goods)• Most of the goods have the following properties:
– rivalry: once consumed, nobody else can consumethe same good (simultaneously, at least)
– excludability: people who do not pay the fair cost canbe prevented from enjoying the benefit of the good
Public goods• Public goods are a particular type of goods which have
both (or part) of the following properties:– non-rivalry: consumption by one person does not
affect the consumption opportunities of the others– non-excludability: once provided, no one can be
prevented from enjoying the benefits
1.1 Private goods, public goods 3
Real-world examples
Pure public goods (almost, I should say)• Common examples of public goods include:
– national defense services– lighthouse and street lighting– environmental services (e.g., clean air)
Impure public goods• Goods in between private and pure public goods:
a) broadcasting services (excludable),b) fishery stock in the high seas (rivalrous),c) national parks and roads (partially rivalrous,
excludable at some cost),where goods of type a) is called club goods while b) iscalled common-pool (or -property) resources
1.2 Real-world examples 4
Why public goods?
Another type of market failure• You can enjoy the benefit of public goods without
purchasing them (once purchased by somebody else)• Who in the world would pay for public goods then?• Incentive to supply public goods is inevitably weak
since they are not sufficiently paid for→ makes it difficult for public goods to be traded in market
Free-riding and inefficiency• This is what we call the free-riding problem, a
consequence of non-rivalry and non-excludability• Due to free-riding, public goods are likely to be
undersupplied relative to the efficient level• Government, as a result, should play a role
1.3 Why public goods? 5
Simple example
Public broadcasting service• Two consumers, A and B• Strategies: P (pay fee) or N (not pay fee)• High-quality TV program only supplied when both pay• Otherwise low-budget poor-quality program supplied
Payoff matrix• Consider the payoff matrix listed below• Assume 2c > g > c > 0 and find the Nash equilibrium
Consumer BP N
P g − c , g − c g/2 − c , g/2Consumer A
N g/2 , g/2 − c 0 , 0
1.4 Simple example 6
Modelling public goods
Preference• n ≥ 2 consumers in the economy• Utility function is Ui(g, xi), where xi is consumption of
private good whereas g is consumption of public good• i has endowment xi > 0 of private good (numéraire)• xi can be used for contribution ci ∈ [0, xi] for public
good:xi + ci = xi (1)
Technology• Consumers know that public good is produced by
g = G(∑ni=1 ci) (2)
• Cost of producing g units of public goods is hence
C(g) := G−1(g) (3)2.1 Modelling public goods 7
Characterizing efficient allocation
Theorem (Samuelson, 1954)• If an allocation (g∗, x∗1 , . . . , x∗n) is Pareto efficient, then it
must be the case thatn
∑i=1
Uig(g∗, x∗i )
Uix(g∗, x∗i )
= C′(g∗), (4)
which is called the Samuelson condition• In other words, sum of MRS should be equal to MRT
Proof (sketch)• Suppose that (4) does not hold, say LHS > RHS• Then, for sufficiently small ε ∈ R++, there must exist(δ1, . . . , δn) ∈ Rn
++ such that reallocation to(g∗ + ε, x∗1 − δ1, . . . , x∗n − δn) is feasible andPareto-improving
2.2 Characterizing efficient allocation 8
Nash equilibrium
Public goods provision game• Public goods provision involves strategic interaction
among consumers• You would have little incentive to make a contribution ci
if everybody else’s contribution c−i is sufficient• Considered to be a game
Nash equilibrium• Profile (cN
i )ni=1 of contribution is a Nash equilibrium if
ui(cNi , cN
−i) ≥ ui(ci, cN−i) ∀ci ∈ [0, xi] (5)
for all i ∈ {1, 2, . . . , n}, where
ui(ci, c−i) := Ui(G(ci + ∑j =i ci), xi − ci) (6)
2.3 Nash equilibrium 9
Characterizing Nash equilibrium
First-order condition• If (cN
i )ni=1 is an (interior) Nash equilibrium, then it must
be the case thatUi
g(gN , xNi )
Uix(gN , xN
i )= C′(gN), (7)
wheregN := G(∑i cN
i ) and xNi := xi − cN
i (8)
Inefficiency• Notice that (9) implies
n
∑i=1
Uig(gN , xN
i )
Uix(gN , xN
i )> C′(gN), (9)
violating the Samuelson condition (undersupplied)
2.4 Characterizing Nash equilibrium 10
Example
Setup and Nash equilibrium• Ui(g, xi) := γi ln(g) + ln(xi) for some γi > 0
• G(c) := c1/2 and hence C(g) := g2
• At eqm,
γixi − cN
igN = 2gN (10)
• For sufficiently small ε ∈ R++, define δi ∈ R++ by
δi :=
(γi
xi − cNi
gN − 1n
(∑
jγj
xj − cNj
gN − 2gN
))ε (11)
• Then (gN + ε, xN1 − δ1, . . . , xN
n − δn) Pareto-dominatesthe eqm allocation (gN , xN
1 , . . . , xNn )
2.4 Characterizing Nash equilibrium 11
Another example (quasi-linear utility)
Setup and Nash equilibrium• Ui(g, xi) := γi ln(g) + xi for some γi > 0• Assume γ1 > γj for all j = 1
• G(c) := c1/2 and hence C(g) := g2
• In this case, eqm involves corner solution:
cNj = 0 for all j = 1 (12)
andcN
1 = 2−1γ1, gN = G(cN1 ) = (2−1γ1)
1/2 (13)• Only consumer 1, who most highly values the public
good, makes a contribution and everybody elsecompletely free-rides!
• Inefficiency: gN < g∗ := (2−1 ∑j γj)1/2
2.4 Characterizing Nash equilibrium 12
Lindahl mechanism: the idea
Personalized tax system• Reasons why public goods are undersupplied:
– strategic incentive– marginal value of public goods varies across
consumers, which is not taken into account• Remove the strategic interaction (among consumers)
by transforming the game into market-like setting• Personalize the price so that everybody pays a fair
share of the cost
Stark contrast to private goods• Private goods: shared price × personalized quantity• Public goods: personalized price × shared quantity
3.1 Lindahl mechanism: the idea 13
Lindahl mechanism: procedure
Procedure• First government announces a personalized
tax-transfer system (τi, Ti)ni=1
• Consumer i’s ‘demand’ for g is determined by
(gdi (τi, Ti), xd
i (τi, Ti)) ∈ argmax Ui(g, xi) s.t. xi + τig = xi +Ti
• Given the feedback (gdi (τi, Ti))
ni=1 from consumers,
government adjusts (τ∗i , T∗
i )ni=1 in such a way that
gdi (τ
∗i , T∗
i ) = gdj (τ
∗j , T∗
j ) =: g∗ for all i, j (14)
and
∑i τ∗i = C′(g∗), ∑i T∗
i + C(g∗) = ∑i τ∗i g∗ (15)
• This tax-transfer system is called Lindahl mechanism
3.2 Lindahl mechanism: procedure 14
Lindahl equilibrium (quasi-linear utility)
Computing Lindahl tax rate• Ui(g, xi) := γi ln(g) + xi for some γi > 0
• G(c) := c1/2 and hence C(g) := g2
• Demand function gdi (τi) is
gdi (τi) = γi/τi (16)
• Set τ∗i and T∗
i by
τ∗i := γi(2−1 ∑j γj)
−1/2, T∗i := 1
2n (∑j γj) (17)
so that gdi (τ
∗i ) = (2−1 ∑j γj)
1/2 =: g∗ for all i
• Efficiency restored b/c
∑i
MRS(g∗, x∗i ) = ∑i
γi/g∗ = 2g∗ = MRT(g∗) (18)
3.3 Lindahl mechanism: example 15
Practical relevance
Vulnerable to ‘cheating’• In the Lindahl mechanism, we essentially ask
consumers to report their preference for public goods• In principle, consumers can tell a lie, misrepresenting
her demand for public goods• Mechanism only works if everybody truthfully reveals
their preference
Any incentive to tell a lie?• Suppose that consumers know how the Lindahl
mechanism works• Then they will realize that they can be (individually)
better off by understating their demand for public goods• Incentive to report gd
i (τi) = γi/τi for some γi < γi
3.4 Practical relevance 16
Alternative to Lindahl mechanism
Mechanism design• In general, a mechanism is said to be
demand-revealing if nobody has an incentive tomisrepresent their demand
• Lindahl mechanism is, by design, not demand-revealing• Is it possible to design a mechanism that is
demand-revealing?
Vickrey-Clarke-Groves mechanism• One example of demand-revealing mechanisms is
Vickrey-Clarke-Groves (VCG) mechanism• Under this mechanism, everybody truthfully reveals
their preference even if they are not required to do so• But this is only achieved at the cost of efficiency
4.1 Alternative to Lindahl mechanism 17
VCG mechanism: procedure
Quasi-linear utility• Consumers’ preference is represented by
Ui(g, xi) := vi(g) + xi (19)
• Here vi(g) is the true valuation of public goods, which isunknown to government
• Special case is vi(g) := γi ln(g)
Procedure• Government first asks consumers to reveal vi : R+ → R• Consumers report vi : R+ → R (can be different from vi)• Given the feedback v := (vi)
Ni=1, government then
decides the level g(v) of public goods as well as thepersonalized lump-sum tax ci(v)
4.2 VCG mechanism: procedure 18
VCG mechanism: design
Designing the mechanism• Prior to asking consumers to reveal their preference,
government announces that g(v) is determined by
g(v) ∈ argmaxg
(N
∑i=1
vi(g)− C(g)
)(20)
and ci(v) is determined by
ci(v) := ti(v)−(
∑j =i
vj(g(v))− C(g(v))
)(21)
for some ti(v) ∈ R (which can depend on v := (vi)Ni=1)
• Consumer i’s utility is then
ui(vi, v−i) := vi(g(v)) + xi − ci(v) (22)
4.3 VCG mechanism: design 19
VCG mechanism: Nash equilibrium
Truth-telling is a dominant strategy• Suppose that everybody else truthfully reveals their
preference, i.e.,vj = vj ∀j = i (23)
• Then, consumer i’s utility is
ui(vi, v−i) = vi(g(vi, v−i)) + xi − ci(vi, v−i) (24)∝ vi(g(vi, v−i)) + ∑
j =ivj(g(vi, v−i))− C(g(vi, v−i)),
where
g(vi, v−i) ∈ argmaxg
(vi(g) + ∑j =i vj(g)− C(g)
)(25)
• Reporting vi = vi (truth-telling) is the best response!
4.4 VCG mechanism: Nash equilibrium 20
VCG mechanism: feasibility
Designing the lump-sum tax• Mechanism is only feasible if ∑N
i=1 ci(v) ≥ C(g(v)) forany v := (vi)
Ni=1, which is equivalent to
N
∑i=1
ti(v) ≥ (N − 1)
(N
∑i=1
vi(g(v))− C(g(v))
)(26)
orN
∑i=1
{ti(v)−
(vi(g(v))− N − 1
NC(g(v))
)}≥ 0 (27)
• Feasibility guaranteed by setting
ti(v) := maxg
(∑j =i
vj(g)− N − 1N
C(g)
)(28)
4.5 VCG mechanism: feasibility 21
VCG mechanism: inefficiency
Government’s budget surplus• VCG mechanism successfully encourages people to
reveal their preference• Government’s budget, on the other hand, is not
balanced in general• Budget surplus: ∑N
i=1 ci(v)− C(g(v)) ≥ 0
Cost of information• Surplus cannot be transferred back to consumers
because that would provide an incentive to tell a lie• Surplus, if any, has to be wasted by design• Pareto efficiency needs to be given up if we want to
make the mechanism demand-revealing• Can be interpreted as the cost of information
4.6 VCG mechanism: inefficiency 22