Professional Documents
Culture Documents
Prelude
Expenditure on private goods – determined by prices
2
Prelude
Decisions about resource allocations in the public sector are quite
differently made
3
Prelude
Thus, there is a major difference between how an individual decides to
spend his or her own money and Government decides to spend the
public money
The vote of a member of Parliament is supposed to reflect the viewsof
constituents
4
Questions to start with
In what ways does collective decision making, such as determining
the level of public goods, differ from standard decision making within
a household?
When individuals differ in what they want, say, about the level of
expenditures on a public good, how are those differences resolved?
5
The Problem of Preference Revelation
Individuals’ desirability of one private good versus another is revealed
by a simple action—by buying the good or not
6
The Problem of Preference Revelation
Election results are quite confusing – some polls suggest that voters
would be willing to pay higher taxes or accept expenditure cuts to
reduce the deficit, but other polls suggest otherwise!
Example from the US: Polls voters consistently say they believe that
the government should spend less on assistance to foreign countries,
when asked how much should be spent, they give a number
considerably in excess of what the United States is currently spending
Only some concrete trade-offs might help them think more precisely
8
Individual Preferences for Public Goods
Collective decision making is difficult because different individuals
have different views
9
Individual Preferences for Public Goods
In the case of private goods, rich and poor individuals typically pay the
same price; with public goods, in effect, richer individuals typically
have to pay a higher price
The tax price is the additional amount an individual must pay when
government expenditures increase by one dollar.
A higher tax price by itself means that richer individuals would want a
lower level of expenditures on public goods
10
Individual Preferences for Public Goods
Income effect leading to a higher desired demand; price effect leading
to a lower desired demand – the net effect is ambiguous
Assume there are N people and each must pay the same amount,
regardless of income (uniform taxation) – tax price is just 1/N and
the tax payment is G/N
11
Individual Preferences for Public Goods
If government expenditures increase by a dollar, the individual’s
incremental tax—the tax price—is just Yi /NY.
Thus, an individual with average income (Yi = Y) faces a tax price of 1/N
An individual with above-average income (Yi > Y) faces a higher tax price
An individual with below-average income faces a lower tax price.
12
Individual Preferences for Public Goods
Given the individual’s tax price, we can derive his or her preferred
level of public goods expenditure
14
Individual Preferences for Public Goods
Different individuals will differ with respect to their preferred level of
expenditures.
16
Individual Preferences for Public Goods
With progressive taxation, lower-income individuals will face a
lower tax price than with proportional taxation, so their preferred
level of expenditures will be even higher than with proportional
taxation.
17
Individual Preferences for Public Goods
18
Individual Preferences for Public Goods
Relationship between the level of utility and the level of public goods
expenditure for three different groups—the rich, the poor, and the
“middle”
20
The Problem of Aggregating Preferences
In the public sector decisions are made collectively.
What “the people” want - how can a social decision be made from
these divergent views?
In a dictatorship, the answer is easy: the dictator’s preferences
dominate.
21
The Problem of Aggregating Preferences
There is no such easy resolution in a democracy
Unanimity voting,
Of these, perhaps the most widely employed rule for decision making
in a democracy is simple majority voting.
22
Majority Voting and the Voting Paradox
A majority voting equilibrium requires that there is one alternative
that can win a majority in a contest against any alternative
Voter 1 prefers A to B to C.
Voter 2 prefers C to A to B.
Voter 3 prefers B to C to A.
23
Majority Voting and the Voting Paradox
Assume we vote on A versus B.
Voters 1 and 2 vote for A, so A wins.
Now we vote on A versus C. Voters 2 and 3 prefer C to A, so C wins.
It appears that C should be the social choice. C wins against A, which
wins against B.
But let us now have a direct confrontation between C and B. Both
Voter 1 and Voter 3 prefer B to C.
24
Majority Voting and the Voting Paradox
Assume we began by saying we were going to first Voter 1
vote on B versus C, and put the winner against A. prefers
B beats C, and then A beats B. A to B to C.
But just to check that we had made the right decision
(A), we decide to put A against C. Voter 2
C beats A. So we think C is the winner. prefers
But then we check that by challenging C with B. C to A to B.
B beats C—which was our original vote.
B again appears to be the winner. Voter 3
But just to check, we again challenge it with A. A prefers
again beats B, as we knew from our earlier vote. B to C to A.
The voting process thus goes on and on …..
25
Majority Voting and the Voting Paradox
Often, to avoid these voting cycles, democracies organize their
decision making as a sequence of votes
26
Majority Voting and the Voting Paradox
Note, too, that if individuals realize there is going to be a particular
sequence of votes, they may wish to vote strategically.
That is, in the first round of the vote, Voter 1 may not vote his or her
true preferences on, say, A versus B, but think through the consequences
of that for the eventual equilibrium.
The voter may vote for B, even though he or she would prefer A,
knowing that in a contest between C and B, B will win, whereas in a
contest between A and C, C might win.
27
Majority Voting and the Voting Paradox
This analysis leads to two questions –
1. Are there voting rules that will ensure a determinate outcome for
any vote?
2. Are there any circumstances under which simple majority voting
will yield a determinate outcome?
It turns out that the voting paradox cannot be resolved through voting
rules, but there are indeed circumstances in which majority voting
yields clear decisions.
28
Arrow’s Impossibility Theorem
An endless cycle of voting is clearly an unsatisfactory state of affairs.
It is natural to ask, then, whether there is any other political
mechanism (any other set of rules for making social decisions) that
eliminates this problem.
In looking for a system that would satisfy all four of these properties, a
number of alternative rules have been examined, but each fails one or
more of the requirements.
30
Arrow’s Impossibility Theorem
For instance, rank-order voting (individuals rank the alternatives,
then the ranks assigned by all individuals are added together, and the
alternative with the lowest score wins) does not satisfy the
“independence of irrelevant alternatives” criterion.
32
Notations and Pareto Criteria
𝑅𝑅𝑖𝑖 is person i's preference relation: implying 𝑥𝑥𝑥𝑥𝑖𝑖 𝑦𝑦 means 𝑖𝑖 likes 𝑥𝑥 as
well as 𝑦𝑦.
𝑃𝑃𝑖𝑖 is the associated strict preference relation
𝐼𝐼𝑖𝑖 is the associated indifference relation
Also, 𝑢𝑢𝑖𝑖 is person i's utility function
33
Characteristics
Completeness and transitivity: either xRy or yRx must hold and
xRy and yRz must imply xRz.
Majority voting gives non-transitive social rankings.
34
Characteristics
Independence of irrelevant alternatives: If people's feelings
change about some set of irrelevant alternatives, but do not change
about the pair of alternatives x and y, then an Arrow social welfare
function must preserve the social ordering of x and y.
35
Example
Two individuals (1, 2) and three alternatives (x, y, z)
xPi y ⇒ i prefers x to y
Given 3 alternatives, there are only six ways individual 1 can order the
alternatives
36
Example
He can prefer x to y to z, or he can prefer x to z to y, and so on…
Each cell in this table shows a possible pair of rankings of the three
alternatives by individuals 1 and 2
37
Preference Profiles
38
Arrow social welfare function
Our concern here is whether or not there is a foolproof rule to transform
any cell in the table into a social preference relation.
Let R stand for a social preference relation, so xRy means x is socially at least
as good as y.
41
Applying the requirements
Now let’s apply the condition of Independence of irrelevant
alternatives
Then independence requires that xPy hold whenever xP1 y and yP2 x
no matter how 1 and 2 rank alternative z.
42
Applying the requirements
Similarly, if yPx (or xIy) holds when person 1 prefers y to x to z and
person 2 prefers x to y to z, then yPx (or xIy) must hold whenever
yP1 x and xP2 y
43
Applying the requirements
Let’s consider the cells again such that we can indicate them in terms
of the following preferences –
if xPy ( xP1 y and yP2 x), cell is marked with ×
if yPx ( yP1 x and xP2 y ), cell is marked with 0
Similarly we can indicate the social rankings over x-z and y-z
44
Applying the requirements
The crossed cells all produce the same x-y social rankings. The circled
cells all produce the same x-y social rankings (which need not be the
same as in the crossed cells).
45
Applying the requirements
The crossed cells all produce the same x-z social rankings. The circled
cells all produce the same x-z social rankings (which need not be the
same as in the crossed cells).
46
Applying the requirements
The crossed cells all produce the same y-z social rankings. The circled
cells all produce the same y-z social rankings (which need not be the
same as in the crossed cells).
47
Arrow's Impossibility Theorem
Does there exist a foolproof rule for discovering, or for defining,
social preferences?
48
Arrow's Impossibility Theorem
We start by looking at the preference profile of the first row, second
column cell of the first table
There are three and only three complete and transitive social
preference orderings which satisfy xPy and xPz.
49
Case 1: yPz
If yPz holds in the first row, second column cell, then independence
requires that y be socially preferred to z whenever individual
preferences about y and z are the same as they are in that cell.
50
Case 1: yPz
Now consider the first row, fifth column cell, or cell number 2 in the
last table.
Pareto consistency requires that xPy here, but xPy and yPz implies xPz,
by transitivity.
Therefore, xPz holds in all the cells indicated in the following table
51
Case 1: yPz
52
Case 1: yPz
Now we have xPz in cell 3.
But this allows us to use independence again to fill in eight more bits
of information
When done, the result is the pattern of social preferences will be like -
53
Case 1: yPz
54
Case 1: yPz
But the social preferences shown in the last table are identical to
person 1's preferences.
Either 1 or 2 is a dictator
55
What we have already learned
The Problem of Preference Revelation – reveal truthfully their
preferences concerning public goods
56
What we have already learned
What “the people” want - how can a social decision be made from
these divergent views?
57
What we have already learned
Arrow’s Impossibility Theorem
Preference profiles
58
Voting System
Arrow established that there does not exist any way of adding together
the preferences of different individuals to satisfy all the desired
characteristics of a choice mechanism
Similarly, there does not, in general, exist any voting system in which
individuals will always vote their true preferences
59
Single-peaked Preferences
Arrow established that there is no voting rule that always satisfies the
five desirable properties of a social choice mechanism
Let’s reconsider individual the utility curves (as a function of the level
of expenditure on public goods)
There, each individual has a single peak to his or her preference profile
- single-peaked preferences
60
Single-peaked Preferences
This property of single-peaked preferences is enough to
guarantee the existence of a majority voting equilibrium.
61
Double-peaked Preferences
Example of double-peaked preference
62
Double-peaked Preferences
Why double peaked?
63
Single-peaked Preferences
Preferences for a single public good (with no private good option, unlike
education) are usually single-peaked,
64
Single-peaked Preferences
This can be seen most clearly in considering the structure of income
taxation.
Suppose we are voting among three income tax schedules that are
designed to raise the same amount of revenue.
Let’s assume that there are three groups of individuals with equal
numbers—the poor, the middle class, and the rich
65
Single-peaked Preferences
The three tax schedules are denoted as A, B, and C
The poor and the middle class then get together and propose tax
schedule B
66
Single-peaked Preferences
The three tax schedules are denoted as A, B, and C
This reduces the taxes they pay but taxes the rich much more heavily.
67
Single-peaked Preferences
Now, the rich propose to the poor: “Because you are more needy, why
don’t we lower your taxes somewhat more? At the same time, we’ll
adjust the tax schedule at the upper end, to reduce the inequities
associated with excessive taxation.”
68
Single-peaked Preferences
69
The Median Voter
When preferences are single-peaked, we asserted that there is a well-
defined majority voting equilibrium.
The median individual is the individual such that half prefer less and
half prefer more than he or she does
70
The Median Voter
Consider the preferred levels of expenditure (in $) on public goods
Lucy Tom Jim John Jill
600 800 1000 1200 1400
71
The Median Voter
If any level of expenditure below $1000 is voted on against $1000, Jim
plus all of those who want more than $1000 vote for $1000
Because Jim is the median voter, a majority cast their vote for $1000.
72
The Inefficiency of the Majority Voting
Equilibrium
The median voter determines the level of expenditure on public goods
The median voter’s benefits are lower than total social benefits (which
includes all the benefits that accrue to others), but so are his or her
costs.
73
The Inefficiency of the Majority Voting
Equilibrium
Whether there is too much or too little expenditure on public goods
thus depends on whether the median voter’s share of total (marginal)
costs is less or greater than his or her share of total (marginal)
benefits.
With uniform taxation, the tax price for each individual would be 1/N
74
The Inefficiency of the Majority Voting
Equilibrium
With proportional taxation, it would be Ym/YN, where Ym is the
income of the median voter and Y is average income
However, most income distributions are skewed – there are a few very
rich individuals, which increases the average income
75
The Inefficiency of the Majority Voting
Equilibrium
As a result, average income exceeds the income of the median
individual (Ym<Y) so that with proportional taxation the tax price is
even less than 1/N.
76
The Inefficiency of the Majority Voting
Equilibrium
What fraction of the total marginal benefits accrues to the median
voter?
If the marginal benefits are uniform, then the median voter gets 1/N
of the total marginal benefits.
So, with uniform taxation, the median voter would get 1/N of the
total social benefits and bear 1/N of the total costs.
77
The Inefficiency of the Majority Voting
Equilibrium
With proportional or progressive taxation, however, the median
voter’s share of the cost would be smaller than his or her share of the
benefits, and he or she would vote for excessive expenditures—that is,
for a level of expenditure at which the sum of the marginal benefits is
less than the total marginal cost to society.
78
The Inefficiency of the Majority Voting
Equilibrium
According to the median voter theory, there is an especially large
tendency for an oversupply of such goods.
79
Two-party System and Majority Voting
Equilibrium
An elected representative bears a negligible fraction of the costs of,
and receives a negligible fraction of the benefits from, an increase in
government expenditure.
What can economic theory say about how he or she should vote?
80
Two-party System and Majority Voting
Equilibrium
Assume there are two parties, “R” and “D.” Party R takes the position
of Party D as a given.
81
Two-party System and Majority Voting
Equilibrium
Let Gm be the preferred level of expenditure of the median voter.
Suppose Party D chooses GD >Gm.
Then, if Party R takes a position between Gm and GD, it will get all the
voters who prefer an expenditure level less than or equal to Gm, and
some who prefer slightly more; thus Party R gets over 50 percent of
the votes and wins.
82
Two-party System and Majority Voting
Equilibrium
In response, Party D will choose a position, G’D, between Gm and GR,
which wins against GR.
Then, however, Party R chooses a position, G’R, between G’D and Gm.
The process continues until both parties stand for the same position:
that of the median voter (Gm).
83
Two-party System and Majority Voting
Equilibrium
This result is consistent with the widely observed allegation that with
our two-party system, voters get no choice: both parties take a
“middle-of-the-road” position
Limitations –
There may not be a majority voting equilibrium. Equilibrium exists if
individuals have single-peaked preferences (single dimension). Multi-
dimension may lead to the problem of cyclicity
Participation in the political process (costs of being informed and
voting)
84
Two-party System and Majority Voting
Equilibrium
In particular, voters whose preferences are near the median have little
incentive to be politically active, particularly if they believe that the
political process will reflect their preferences anyway .
Thus, it may be in the interests of those who are more extreme to
attempt to pull their party away from the center.
This tendency for greater political activism at the extremes may
partially off set the median-directed tendencies.
There is a tension between the centripetal force of the median voter
model and the centrifugal force of the base voter mobilization model.
85
Majority Voting – Summary
Proposal that gets a majority against all other wins.
86
Any alternative?
Majority voting – quite a few problems
87
Lindahl Equilibrium
Erik Lindahl came up with the famous alternative known as the
Lindahl solution.
It draws heavily from the way that the market works in providing
private goods
88
Lindahl Equilibrium
For public goods, efficient level that can be provided can be
interpreted as – intersection of the “collective’’ demand curve (formed
by adding vertically each individual’s demand curve) with the supply
curve
The demand curves are generated by asking the individual how much
of the public good he or she would demand if he or she were to pay so
much for each unit produced
89
Lindahl Equilibrium
For public goods, efficient level that can be provided can be
interpreted as – intersection of the “collective’’ demand curve (formed
by adding vertically each individual’s demand curve) with the supply
curve
The demand curves are generated by asking the individual how much
of the public good he or she would demand if he or she were to pay so
much for each unit produced
90
Individual 1
91
Individual 2
92
Lindahl Equilibrium
Vertical addition of demand curves and intersection with the supply
curve leads to Lindahl equilibrium
At G*, the sum of the marginal benefits (the total marginal willingness
to pay) just equals the marginal cost.
93
Lindahl Equilibrium
94
Lindahl Equilibrium
The Lindahl equilibrium is thus a set of tax prices (the amount each
individual has to pay if one more unit of the public good is produced)
adding up to the marginal cost of production, such that, given those
tax prices, every individual prefers the same level f expenditures, G*.
95
Lindahl Equilibrium
Individuals who are disadvantaged by this particular Pareto efficient
point will not agree to the use of this mechanism for determining the
allocation of public goods
The demand curves that are used in the Lindahl analysis were drawn
under the hypothesis that individuals face a given tax price; they
believe that nothing they say will alter what they have to pay per unit
of public expenditure.
97
Comparison
Majority voting
Equilibrium may not exist.
When equilibrium exists, in general, it is not Pareto efficient.
Lindahl equilibrium
Equilibrium always exists.
Equilibrium is always Pareto efficient.
Individuals do not have incentive for truthful revelation of
preferences.
98
A Simple Model
A good is public if it is by nature available to all: if one man uses it,
everyone can use it.
99
A Simple Model
The answer to the question depends both on people's demand for it,
and on the nature of the productive sector of the economy.
We assume that there are only two goods, one private, and one public.
100
A Simple Model
We assume that person i's utility is the sum of the quantity of the
private good he consumes, plus a well-behaved function of the
quantity of the public good produced and available to all, including i.
Note that x can be viewed as the quantity (or size) of the public good
in physical units, or in dollars, since we assume that the prices of both
the public and the private good are one
101
A Simple Model
yi = person i's quantity of the private good
That is, i's utility is the sum of the function vi, which depends only on
X, plus i's quantity of the private good.
MU of the public good for i, when the quantity of the public good is x,
vi' ( x)
102
A Simple Model
We also assume that each person starts with an initial endowment of
the private good. ωi
Feasibility condition:
103
Wicksell-Lindahl Tax System
A central government authority called the Public Good Board decides
on the output of the public good X, and collects taxes to pay or X.
Requirements –
it should somehow provide for a Pareto optimal output of the public
good
it ought to link a person's taxes to the benefits he receives
Why should a person who gets little or no utility from the public good
pay the same taxes as a person who gets lots of utility from it?
104
Wicksell-Lindahl Tax System
Let Ti stand for person i's total tax payment to the Public Good Board.
105
Wicksell-Lindahl Tax System
Since, ti is the “share”, we must have ∑t i =1
max ui = vi ( x) + yi
s.t. yi + ti x = ωi
Re-write: max vi ( x) − ti x + ωi
Hence, i's desired quantity of the public good xˆi (ti ) will be
determined by vi' ( x) = ti
106
Wicksell-Lindahl Tax System
Observation – the higher is person i's share ( ti ), the lower is the
quantity of the public good xˆi (ti ) he wants produced (analogous to
law of demand)
If the actual output of the public good happens to coincide with i's
desired output of the public good, that is, x = xˆi (ti ) , then, then i's
marginal utility from the public good vi' ( x) equals his tax share ti
But how can the actual output be made to agree with i's desired
output?
107
Wicksell-Lindahl Tax System
How can the Public Good Board find a way to insure that each and
every person's tax share is linked to his marginal benefit?
The trick of the W-L tax scheme is to adjust the tax shares until every
person agrees on the desired output of the public good.
108
Wicksell-Lindahl Tax System
Solution: decrease t1 and increase t 2
109
Wicksell-Lindahl Tax System
Formally, a Lindahl equilibrium is defined to be a vector of tax shares
(t1 , t 2 ,..., t n ) and a level of output x for the public good, such that,
for all i, when i's tax share is ti his desired level of public good output
equals x.
110
Wicksell-Lindahl Tax System
It should be emphasized that the connection is with marginal utility
rather than total utility.
Now, ∑t i =1
111
Wicksell-Lindahl Tax System
Drawbacks of the W-L tax system (or the Lindahl equilibrium)
2. Excludability
112
Wicksell-Lindahl Tax System
Problem of Incentives: When the Public Good Board calculates the
Lindahl equilibrium, when it calculates the appropriate vector of tax
shares (t1 , t 2 ,..., t n ) and the output for the public good X, it relies on
information it receives from the individuals.
It needs each individual's xˆi (ti ) schedule, which we might call i's
demand schedule for the public good.
113
Wicksell-Lindahl Tax System
So, when n is large, lying will induce the Board to produce a slightly
smaller x.
So the person who is willing to conceal his demand for the public good
will end up with almost as much of it as he would have otherwise, and
he'll pay a lot less per unit than the would have otherwise!
114
Fixed Tax Shares
Fixed tax share, median voter, majority voting
ti = 1/n; T=n.ti
ti = wi/sum wi
ti, x
Under demand revealing tax system the Government sets total tax Ti
instead of ti following the rule –
Ti = xˆ − ∑ v j ( xˆ ) j≠i
j
Let’s consider the ith individual. He doesn’t know whether v−' i is true
or not
117
The Demand-Revealing Tax Scheme
Individual i (and everyone else) knows that Ti = xˆ − ∑ v j ( xˆ )
j
Therefore, max vi ( xˆ ) − [ xˆ − ∑ v j ( xˆ )]
j
118
The Demand-Revealing Tax Scheme
Individual i knows that if i truly reports ( vi ), Board comes up with x̂
such that
max ∑ vi ( x) − x = vi ( x) + ∑ v j ( x) − x = vi ( x) − [ x − ∑ v j ( x)]
i≠ j i≠ j
Lying is sub-optimal
119
The Demand-Revealing Tax Scheme
Disadvantage – insufficient funds.
But we can’t have a tax system such that (i) truth, (ii) Pareto
optimality and (iii) non-wastage ( ∑T i = xˆ )
120
References
Stiglitz, JE and JK Rosengard (2015). Economics of the Public Sector,
4th Edn. WW Norton. Chapter 9
121