Transcript Document
Frank Cowell: UB Public Economics
June 2005
Optimal Tax Design
Public Economics: University of Barcelona
Frank Cowell
http://darp.lse.ac.uk/ub
Frank Cowell:
Purpose of tax design
UB Public Economics
The issue of design is fundamental to public economics
Move from what we would like to achieve…
…to what we can actually implement
Plenty of examples of this issue:
Public-good provision
Regulation
Social insurance
Optimal taxation – see below.
Important to be clear what the purpose of the tax design
problem is.
A brief review of the elements of the problem. …
Frank Cowell:
Components of the problem
UB Public Economics
Objectives
Scope for policy
Could be an attempt to satisfy a particular objective
function or class of functions
Could be a characterisation of policies that achieve
some broad objectives.
Methods of intervention
Constraints
Informational problems
Available tools
The tax base
Direct and indirect taxation
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
Why this kind of
problem is set up
General labour
model
“linear” labour
model
Education
model
Generalisations
•Objectives
•Scope for policy
•Informational issues
•Available tools
Frank Cowell:
Specific objectives?
UB Public Economics
Could be a class of
The objectives of the tax design
could include:
functions
Could be
incorporated in
1.
Bergson-Samuelson welfare maximisation
objective #1
2.
Overall concern for efficiency
3.
Overall concern for reduction of inequality
of outcome.
4.
Inequality of opportunity
5.
Poverty, horizontal inequity...
More than one of the above may be relevant.
Frank Cowell:
Implementation of objectives
UB Public Economics
What is domain of the SWF?
Need a model of
cardinal, comparable
utility
Incomes?
Individual utilities?
Welfarist approach usually founded
on this basis
What social Data
entities?
is often on
this basis…
Individuals
…or this
Families
Household units?
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
Types of
intervention. The
tax base
General labour
model
“linear” labour
model
Education
model
Generalisations
•Objectives
•Scope for policy
•Informational issues
•Available tools
Frank Cowell:
Scope for policy
UB Public Economics
What is potentially achievable?
We need to do this before we can examine
specific policy tools and their associated
constraints.
If we have in mind income redistribution it is
appropriate to look at the determinants of income
Do this within the context of an elementary
microeconomic model.
Frank Cowell:
The Composition of Income
UB Public Economics
Take the standard microeconomic model of a
person’s total income in a market economy
Composed of resources valued at their
market prices:
endowment
income
of good i
Non-market
income
price of
good i
Does this mean public policy has to be limited to .
redistributing resources, or
2. manipulating prices?
There could be other forms of income
Problems with 1 and 2 above are also important
And there may be other types of intervention
1.
Frank Cowell:
Problems with redistributing
resources:
UB Public Economics
The lump-sum tax issue:
Non-transferability
Special information – such as personal characteristics
Political problems of implementation
Fixed resources
Inalienability of certain rights – No slavery
Ways of getting round these problems?
Could redistribute the purchasing power generated by the
resource?
Or modify the supply of “co-operant factors”?
Frank Cowell:
Problems with price manipulation
UB Public Economics
Identification of commodities
Complexity
Proliferation of implied pricing structures
Informational problems
The boundary problem
Artificial definition of a good or service on which a tax
is to be levied.
Uncertainty leads to wrong price signals?
Misinformation leads to wrong price signals?
May even be missing markets
Need to focus on economics of information
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
Fundamental
theoretical issues
in design
problem
General labour
model
“linear” labour
model
Education
model
Generalisations
•Objectives
•Scope for policy
•Informational issues
•Available tools
Frank Cowell:
Informational issues in
microeconomics
UB Public Economics
There are two key types of informational problem:
Both can be relevant to policy design.
Hidden action:
Hidden information:
Regulation and optimal contracts.
Moral hazard in social insurance
Compliance issues.
Problems of “tailoring” tax rates.
Adverse selection in social insurance.
Focus on this issue here
But the “information issue” is quite deep:
There is connection with discussion of social welfare
A fundamental relationship with the “Arrow” problem
Frank Cowell:
Social values: the Arrow problem
UB Public Economics
Uses weak assumptions about preferences/values
Uses a general notion of social preferences
The constitution
A map from set of preference profiles to social preference
Also weak assumptions about the constitution
Well-defined individual orderings over social states
Well-defined social ordering over social states
Universal Domain
Pareto Unanimity
Independence of Irrelevant Alternatives
Non-Dictatorship
There’s no constitution that does all four
Except in cases where there are less than three social states
Frank Cowell:
Social-choice function
UB Public Economics
Similar to the concept of constitution
But maps from set of preference profiles to set of social
states
Not surprising to find result similar to Arrow
Given a particular set of preferences for the population
Picks out the preferred social state
Introduce weak conditions on the Social-choice function
There’s no SCF that satisfies all of them
But key point concerns the implementation issue
Frank Cowell:
Implementation
UB Public Economics
Is the social-choice function consistent with private
economic behaviour?
Yes if the social state picked out by the SCF corresponds
to an equilibrium
Problem becomes finding an appropriate mechanism
mechanism can be thought of as a kind of cut-down game
to be interesting the game is one of imperfect information
is the desired social state an equilibrium of the game?
There is a wide range of possible mechanisms
Focus on a type that is useful for expositional purposes...
Frank Cowell:
Direct mechanisms
UB Public Economics
Map from collection of preferences to states
Here the SCF is the mechanism itself
An SCF that encourages misrepresentation may be of
limited use
Is truthful implementation possible?
Involves a very simple game.
The game is “show me your utility function”
Enables us to focus directly on the informational aspects of
implementation
Will people announce their true attributes?
Will it be a dominant strategy to do so?
Introduce another key result
Frank Cowell:
Gibbard-Satterthwaite result
UB Public Economics
Can be stated in a variety of ways.
A standard versions is:
If the set of social states contains at least three elements;
...and the social choice function is defined for the all logically
possible preference profiles...
...and the SCF is truthfully implementable in dominant strategies...
...then the SCF must be dictatorial
Closely related to the Arrow theorem
Has profound implications for public economics
Misinformation may be endemic to the design problem
May only get truth-telling mechanisms in special cases
Underlies issues of public-good provision, regulation, tax design
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
What practical
options available
to achieve the
objectives?
General labour
model
“linear” labour
model
Education
model
Generalisations
•Objectives
•Scope for policy
•Informational issues
•Available tools
Frank Cowell:
Informational issues in taxation
UB Public Economics
What distinguishes taxation from highway
robbery?
Taxation principles
Appropriate information
What information is/should be available?
Attributes
Behaviour
Frank Cowell:
Available tools
UB Public Economics
Availability determined by a variety of
considerations.
Fundamental economic constraints
Institutional constraints. These may come from:
Legal restrictions
Administrative considerations
Historical precedent
But each of these institutional aspects may really
follow from the economics.
Frank Cowell:
The Tax Base
UB Public Economics
We focus here on the taxation of individuals rather than
corporations or other entities.
An approach to the individual tax-base might begin with
number
an examination
of the individual’s budget constraint:
of goods
expenditure
consumption of
good i
income
So taxation might be based on consumption of specific
goods or on some concept of income or expenditure
We will see that using the above as an elementary
method of classifying taxes can be misleading
First take a closer look at income:
Frank Cowell:
A fundamental difference?
UB Public Economics
It is tempting to think of the distinction between
different types of tax in terms of the budget
Indirect
constraint:
taxes here?
Direct taxes
here?
This misses the point .
Any tax on RHS can be converted to tax on LHS
Real question is about information
Frank Cowell:
Information again
UB Public Economics
The government and its agencies must work with
imperfect information.
To model taxes appropriately need to take this
into account.
Information imposes specific constraints on tax
design
Income
In a typical market economy
there are two main
Total expenditure
types of information:Age, marital status?
About individuals
About transactions
Expenditure by product category
Expenditure by industry
Input and output quantities
Frank Cowell:
Fundamental constraints...
Public budget constraints
UB Public Economics
Participation constraints
Example: In simple redistribution sum of net receipts
(taxes cash subsidies) must be zero
Example: Labour supply
Incentive-compatibility (self-selection) constraints
Example: Differential subsidies for specific
commodities
Frank Cowell:
Design basics: summary
UB Public Economics
Objectives follow on logically from our
discussion in previous lectures.
Beware of oversimplifying assumptions
about the tax base.
Information plays a key role.
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
What types of tax
formula used in
theory and
practice?
General labour
model
“Linear” labour
model
Education
model
Generalisations
•Tax schedules
•Outline of problem
•The solution
Frank Cowell:
Income tax – example of design
problem
UB Public Economics
Standard types of tax
simple examples
integration with income support
General issues of how to set up an optimisation
problem
Solution of optimal tax problem:
Solution of the general tax design problem
Solution of the special “linear” case
Alternative models of optimal income tax
Frank Cowell:
Income tax – notation
UB Public Economics
y – taxable income
c – disposable income (“consumption”)
c(y) = y – T(y)
T(·) – tax schedule
c(·) – disposable income
0 t schedule
1
t – marginal tax rate
y0 – exemption-level income
B – lumpsum benefit/guaranteed income
c=c(y)
UB Public Economics
disposable income
Frank Cowell:
Income space
pre-tax income
y
Frank Cowell:
The simple income tax
c=c(y)
UB Public Economics
Marginal retention
rate
1-t
Exemption
level
y0
y
Frank Cowell:
...extended to Negative Income Tax
c=c(y)
UB Public Economics
Incomes
subsidised
through NIT
1-t
B = t y0
B
y0
y
Frank Cowell:
How to generalise this approach…?
UB Public Economics
Other functional forms of the income tax
Administrative complexity of IT
Interaction with other contingent taxes and
benefits.
Frank Cowell:
Increasing marginal tax rate t
c(y)
UB Public Economics
y
Frank Cowell:
Example 1
UB Public Economics
UK:
piecewise linear tax
stepwise jumps in MTR
compare with contingent tax/benefit model
Germany:
linearly increasing marginal tax rate
quadratic tax and disposable income schedules
Frank Cowell:
Example 2
UB Public Economics
Germany 1981-1985, single person (§32a Einkommensteuergesetz):
up to 4,212: T = 0
4,213 to 18,000: T = 0.22y – 926
18,001 to 59,999: T = 3.05 z4 – 73.76 z3 + 695 z2 + 2,200 z + 3,034
z = y/10,000 - 18,000;
60,000 to 129,999: T = 0.09z4 – 5.45z3 + 88.13 z2 + 5,040 z + 20,018
z = y/10,000 - 60,000;
from 130,000: T = 0.56 y – 14,837
(units: DM)
80,000
70,000
60,000
50,000
40,000
30,000
20,000
10,000
0
0
20000
40000
60000
80000
100000
120000
140000
Frank Cowell:
Interaction with income support
UB Public Economics
c(y)
Straight income
tax at constant
marginal rate
“Clawback”
of support
Tax-payments kick
in with benefits
Untaxed
income
support
B
0
y1
y0
y2
y
Frank Cowell:
The approach to IT – summary
UB Public Economics
The “linear” form may be a reasonable approximation to
some practical cases
We may also see an appealing intuitive argument for
linearity as simplification
“Income tax” may need to be interpreted fairly broadly
Interaction amongst various forms of government
intervention is important for an appropriate model
This may lead to nonlinearity in the effective schedule
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
Basic ingredients
of OIT analysis.
General labour
model
“Linear” labour
model
Education
model
Generalisations
•Tax schedules
•Outline of problem
•The solution
Frank Cowell:
Basic Ingredients of An Optimal
Income Tax model
UB Public Economics
A distribution of abilities
Individuals’ behaviour
Social-welfare function
Feasibility Constraint
Restriction on types of
functional form
What resources are
potentially available
for redistribution?
Frank Cowell:
Distribution of Ability...
Assume...
UB Public Economics
a single source of earning power – “ability”
ability is fully reflected in the (potential) wage w
So ability is effectively measured by w.
the distribution F of w is observable
individual values of w are not observable by the
tax authority
Frank Cowell:
Can we infer the distribution of
ability?
UB Public Economics
Practical approach
Select relevant group or groups in population.
Choose appropriate earnings concept.
male manual workers?
Full time earnings?
Divide earnings by hours to get wages.
Use parametric model to capture shape of
distribution.
Lognormal?
Frank Cowell:
Distribution: example
Example from UK 2000
Gives distribution of y=wh for full-time male manual workers
UB Public Economics
f (y )
NES 2000
Lognormal (5.7,0.13)
y
£5
0
£1
00
£1
50
£2
00
£2
50
£3
00
£3
50
£4
00
£4
50
£5
00
£5
50
£6
00
£6
50
£7
00
£7
50
£8
00
£8
50
£9
00
£9
5
£1 0
,0
00
£0
0
Frank Cowell:
Basic Ingredients of An Optimal
Income Tax model (2)
UB Public Economics
A distribution of abilities
Individuals’ behaviour
Social-welfare function
Feasibility Constraint
Restriction on types of
functional form
In what ways do we
assume that people
will respond to the
tax authority’s
instruments ?
Frank Cowell:
The individual's problem
UB Public Economics
Individual’s utility is determined by disposable
income (consumption) c and leisure.
So the optimisation problem can be written
maxh U(c,h)
subject to c = y – T(y)
and y = wh
This yields maximised utility as a function of
ability (wage):
u(w)
Frank Cowell:
A Characterisation of Tastes
UB Public Economics
Introduce a definition to capture the shape of
individual preferences
Normalised MRS
The following restriction on “regularity” of
The way slope of indifference curve
preferences is important
forability
clean-cut results
changes with
Frank Cowell:
A representation of preferences
UB Public Economics
(leisure)
Frank Cowell:
Indifference curve in (h,c)-space
c
UB Public Economics
(hours worked)
h
Frank Cowell:
Contour translated to (y,c)-space
c
UB Public Economics
slope = q
(gross income)
y
Frank Cowell:
The regularity condition
c
high w
UB Public Economics
low w
Illustrates the qw < 0 property
Ensures “single-crossing” of
ICs for different ability groups
y
Frank Cowell:
Individual's problem: points to
note
UB Public Economics
Incorporates standard assumptions
Same basic model as in earlier lectures
Consistent with the model
y = wh
or with the model
y = wh + I
Frank Cowell:
Basic Ingredients of An Optimal
Income Tax model (3)
UB Public Economics
A distribution of abilities
Individuals’ behaviour
Social-welfare function
Feasibility Constraint
Restriction on types of
functional form
How to represent the
objectives of the
optimisation
problem?
Frank Cowell:
The Government’s Objective...
Take a standard version of the SWF
UB Public Economics
Assume additive separability
Take “weighted average” over types
Maximised utility of a
w-type person
Social evaluation
function
Proportion of w-type persons in
the population
Frank Cowell:
Basic Ingredients of An Optimal
Income Tax model (4)
UB Public Economics
A distribution of abilities
Individuals’ behaviour
Social-welfare function
Feasibility Constraint
Restriction on types of
functional form
Real-world
restrictions on
government and the
design problem
Frank Cowell:
Main types of constraint
UB Public Economics
The Government’s budget constraint
Incentive compatibility
Frank Cowell:
Government’s Budget Constraint
disposable income
in population
Net revenue
requirement
Earnings in
population
UB Public Economics
Frank Cowell:
Two Ability Levels
c
high w
UB Public Economics
c(y)
low w
Incentive compatibility problem:
Original and disposable income
must increase with ability
y
Frank Cowell:
Basic Ingredients of An Optimal
Income Tax model (5)
UB Public Economics
A distribution of abilities
Individuals’ behaviour
Social-welfare function
Feasibility Constraint
Restriction on types of
functional form
Frank Cowell:
Restrictions on form...
UB Public Economics
It may make sense to consider cases where
marginal tax-rate is everywhere constant
Pre-1985 Germany?
(like the NIT model earlier):
Lack of detail
about tails
Administrative costs of general model
Cf the US “flat
tax” discussion
informational problems
“fairness” arguments [?]
Frank Cowell:
Review: ingredients of OIT model
UB Public Economics
A distribution of abilities
Individuals’ behaviour
Assume individualistic additively separable SWF
W = u( u(w) f(w) dw
Feasibility constraints
maxh U(c,h), subject to c = y – T(y) and y = wh
Gives utility as function of ability u(w)
Social-welfare function
Ability is measured by potential wage w.
Distribution F of w is observable
Individual values of w are not observable
The Government’s budget constraint
Incentive compatibility
Restriction on types of functional form
(Piecewise) linear schedule?
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
Characterising a
general optimal
income tax.
General labour
model
“Linear” labour
model
Education
model
Generalisations
•Tax schedules
•Outline of problem
•The solution
Frank Cowell:
The general OIT model
UB Public Economics
No preconditions on form of income tax
Use results from economics of information
Use a general variational approach to give the solution
IC condition required for “sensible” results
However, no consideration of administrative complexity
Has similarities with techniques used for optimal growth
Terminal conditions can be important
Illustrate the variational approach diagrammatically
Use the disposable income schedule c(•)
Frank Cowell:
The general tax-design problem
UB Public Economics
c(y)
Variation in general tax
schedule
y
Frank Cowell:
Individual optimisation
UB Public Economics
Begin with the way each tax-payer is assumed to act.
Disposable
The optimisation
problem can be written
income
work
The function c(·) is chosen by the government
Define normalised MRS
Slope of disposable
income function
The first-order condition is
The solution is of the form u(w) := maxh U(c(wh), h)
Frank Cowell:
Contour in (y,c)-space
c
UB Public Economics
Disposable income
schedule c(•)
optimised
income
Proportional
to work
y
Frank Cowell:
A regularity condition
c
high w
UB Public Economics
low w
If the property qw < 0 holds this
ensures “single-crossing” of ICs for
different ability groups
y
Frank Cowell:
Incentive compatibility condition
c
high w
UB Public Economics
c(y)
low w
Design of c must ensure that
utility and income increase
with ability
y
Frank Cowell:
Incentive compatibility
UB Public Economics
The IC condition means that high ability people should
not have an incentive to “masquerade” as low ability.
This requires maximised utility u(w) to increase in w.
By differentiation of the solution function we have
Optimised
value of h
If c(·) is monotonic and differentiable everywhere
then this becomes
But if these conditions are violated, problems arise…
Frank Cowell:
Fundamental design problem
UB Public Economics
It may seem odd that the IC condition be violated in actual
design
But it can happen by accident:
interaction between income support and income tax.
generated by the desire to “target” support more effectively.
A well-meant gross inefficiency?
Commonly known as
The “notch problem” (US)
The “poverty trap” (UK)
Frank Cowell:
“Notch problem” / “poverty trap”
UB Public Economics
c(y)
Discontinuous nonmonotonic c(·)
Withdrawal of
benefit here
y0
y
Frank Cowell:
Violation of IC condition
UB Public Economics
c(y)
Where high w
would choose
Where high
w “should”
be
low w
y0
y
Frank Cowell:
Government: maximisation
UB Public Economics
Uses a constrained maximum method
But there is a constraint at each ability level from wmin to wmax.
Similar to maximisation over time.
Choose c(·) to max
subject to
and, at each ability level:
disposable income
in population
Net revenue
requirement
Earnings in
population
Frank Cowell:
Government: maximisation
UB Public Economics
Introduce a Lagrange multiplier l for the budget constraint
and a multiplier m(w) for the incentive compatibility
constraint at each ability level.
Then, on rearranging, the Lagrangean is
Frank Cowell:
General Model: Characterisation
of Marginal Tax Rate
UB Public Economics
Lagrange multiplier for
incentive-compatibility
constraint
Lagrange multiplier for
Government budget
constraint
Frank Cowell:
Interpreting the FOC
UB Public Economics
Can be used to give us an impression of the shape
of the solution
But an explicit form for the OIT is usually not
possible
Some key results
First for the overall shape
Second for what happens at each end of the ability
range…
Frank Cowell:
Main result 1
UB Public Economics
Mirrlees 1971: The optimal marginal tax rate must be
greater than or equal to 0 and less than 1
http://darp.lse.ac.uk/papersdb/Mirrlees_(REStud_71).pdf
The condition “0” means that in trying@to raise tax it
never makes sense to introduce a distortionary labour
subsidy − see Tuomala (1990)
The condition “<1” follows from
Agent monotonicity implies
So it is immediate that T'(y) < 1.
For the lower extreme of the distribution need to
look at “bunching”…
Frank Cowell:
Bunching: w'<w''<w'''
UB Public Economics
w'''
w''
c
w'
y
Frank Cowell:
No bunching: w'<w''<w'''
UB Public Economics
c
w'''
w'
w''
y
Frank Cowell:
Main result 2
UB Public Economics
Seade 1977,Ebert 1992. The optimal marginal tax rate:
is 0 on the highest income
is 0 on the lowest income if there is no bunching
is positive on the lowest income if there is bunching
For bottom of distribution see Tuomala (1990)
For top of distribution note the FOC:
At wmax IC constraint becomes irrelevant; so m(wmax) = 0.
Therefore T'(ymax) = T'(wmax h(wmax)) = 0
Frank Cowell:
Problems of the general model (1)
UB Public Economics
There appear to be commonsense general results
And clear-cut results for the extremes,
But little guidance on the structure for the majority of the
workforce.
Some broad principles can be adduced from the first order
conditions.
But you cannot get further without an explicit model
On the distribution of w
On individual preferences
On the SWF
Frank Cowell:
Problems of the general model (2)
UB Public Economics
The results for the extremes are not robust
Should have low or decreasing tax rates close to the top of
the income distribution?
This does not seem to be the case from simulation study
Tuomala: J. Pub. Econ 1984
Part of the problem arises from assumed F(•) of w
convenient to assume that support of the distribution F is finite
But this means an artificial assumption about known “endpoints”
Frank Cowell:
Problems of the general model (3)
UB Public Economics
Most applied models assume something like lognormal
or Pareto
Support is unbounded above.
No “maximum” income
If you rework the model with a distribution that is
“open-ended” at the top things appear very different.
Diamond (1998) uses Pareto.
Gets high marginal tax rates where ability follows a Pareto
distribution http://darp.lse.ac.uk/papersdb/Diamond_(AER98).pdf
Saez (2001) is a general extension of the Mirrlees results
http://darp.lse.ac.uk/papersdb/Saez_(REStud01).pdf
Frank Cowell:
Problems of the general model (4)
UB Public Economics
Cannot get stronger results on tax rates analytically.
But to do this you need to implement a specific model
which can be:
Can do this for special cases
Example Salanie model with quasi-linear preferences
Or could use simulation in a numerical model
Computationally messy
Sensitive to specific assumptions made about labour supply and
ability
It may make sense to impose more structure a priori on the
tax function
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
A “cut-down”
version of the
labour-leisure
problem
General labour
model
“linear” labour
model
Education
model
Generalisations
Frank Cowell:
Approach 2: The linear model
UB Public Economics
Same behavioural assumptions as before
Same objectives
Restriction to linear (affine) tax functions: two parameters
First analysed by Sheshinski (1972)
http://darp.lse.ac.uk/papersdb/Sheshinski_(REStud_72).pdf
Frank Cowell:
Linear Model: outline
UB Public Economics
No longer choosing a general tax/disposable income
schedule c(•)
Instead, just a two-parameter model.
Disposable income is
Marginal
tax rate
Pre-tax
income
Minimum
disposable
income
Simplified version is much more tractable analytically
Frank Cowell:
Arguments for “linear” model
UB Public Economics
Relatively easy to interpret parameters
Pragmatic:
t as uniform marginal tax rate
B as minimum income, or…
B / t as exemption rate
Approximates several countries’ tax systems
Example – piecewise linear tax in UK
Sidesteps the incentive compatibility constraint…
Frank Cowell:
Incentive compatibility resolved
c
high w
UB Public Economics
c(y)
low w
Original and disposable income will
increase with ability
y
Frank Cowell:
Linear Model: Constraint
UB Public Economics
Given that the IC condition vanishes, there is only one
constraint
Government
TaxThe
raised
on working Budget constraint:
Extra revenue
population
Minimum guaranteed
income for all
requirement
In effect this makes the issue a one-variable problem…
Frank Cowell:
Take the linear income tax...
c(y)
UB Public Economics
1-t
B
Note
Marginal tax-rate is constant: t
If B>0 average tax-rate t–B/y is
everywhere rising with income
y
Frank Cowell:
Higher B needs higher t
c(y)
UB Public Economics
1-tDt
B+DB
y
Frank Cowell:
Linear Model: Lagrangean
UB Public Economics
The constrained optimisation problem can be set up as
Lagrange
Social-welfare
the Lagrangean:
multiplier
function
Government budget
constraint
Maximise Lagrangean by choice of tax instruments t
and B
This can be done using classical optimisation methods.
Frank Cowell:
Linear Model: FOC (1)
Maximised
utility
$1
lump-sum
UB Public Economics
Consider the social value of
This is defined as:
Average
social value
Differentiating the Lagrangean with
respect
to B:of $1
should be 1
income.
Frank Cowell:
Linear Model: FOC (2)
UB Public Economics
Differentiating the Lagrangean with respect to t and
Covariance of social marginal
rearranging we get:
valuation and income
Optimal marginal tax rate
Compensated laboursupply elasticity
Again the formula can be used to give guidance on
policy…
Frank Cowell:
Outcomes from the linear model
UB Public Economics
If R = 0 then B > 0
Implies progressive taxation.
FOC cannot be solved to give an explicit formula
The covariance and the elasticities will themselves be functions of
t.
However the “natural” restriction imposed by linearity
makes construction of simulation easier
Better behaved at special points of the distribution
Frank Cowell:
Components of simulation
UB Public Economics
Structure of ability (wage) distribution
Individual preferences
Determines labour supply responses
Social welfare function
Empirically determined?
Use evidence from social surveys etc?
Government budget constraint
Experiment with alternative assumptions
Frank Cowell:
Broome’s revelation
UB Public Economics
John Broome suggested a great simplification for OIT.
t* = 58.6% !!
http://darp.lse.ac.uk/papersdb/Broome_(REStud_75).pdf
The basis for this astounding claim?
When we spot that the tax rate is in fact 2 – 2 the
remark is not so outlandish
Rather it serves as a useful lesson in applied modelling
Frank Cowell:
Broome’s model…
UB Public Economics
But in UK 2000:
Make the (empirically relevant?)
assumption
that£10.53
no-one
1 Average
wage was
/ has
ability less than 0.7071 times thehour
average:
2 Min wage was £4.10!
Take standard Cobb-Douglas preferences:
“Rawlsian” max-min social welfare:
Balanced budget:
Frank Cowell:
The Stern simulation model
UB Public Economics
Stern’s (1976) model is less tongue-in-cheek.
But can be taken as a generalisation of Broome.
Also based on a linear OIT
Ingredients are:
Lognormal ability
Isoelastic utility
Isoelastic social welfare
A variety of assumptions about the government budget constraint
Frank Cowell:
Representation of ability
distribution
UB Public Economics
Simple two parameter distribution
L(w; m, s2 )
First parameter m is log of the median
The second parameter s2 is itself an inequality index – the variance
of log income.
Support is [0, )
Not a bad approximation to empirical distributions
Particularly for manual workers
Stern assumed s = 0.39 (same as Mirrlees)
In this case less than 2% of the population have less than 0.7071 ×
mean (Broome)
Frank Cowell:
The lognormal distribution
f(w)
UB Public Economics
—L(w; 0, 0.25 )
…L(w; 0, 1.0 )
0
0
1
2
3
4
w
Frank Cowell:
Isoelastic utility
UB Public Economics
Take an empirically relevant version of household
elasticity of
utility:
Substitution ( 0)
hours worked
consumption
Becomes the Broome model in the case s=1
Frank Cowell:
Labour Supply and Income...
Define q := [1t] w
UB Public Economics
Could have backward-bending
labour supply if s<1
Frank Cowell:
Resulting Labour Supply and
Income… (Broome case)
UB Public Economics
Frank Cowell:
Standard SWF
Take additive form of Bergson-Samuelson SWF:
UB Public Economics
W = u(u) dF(u)
= u(u) f(u) du
Use the iso-elastic form of the (social) u-function:
u 1–e – 1
u(u) = ————, e
1–e
Bentham corresponds to the case e=.
Max-min (“Rawls”) corresponds to the case e=.
Frank Cowell:
Stern's Optimal Income Tax
Rates
UB Public Economics
s
e=
e= 1
e=
0.2
0.4
0.6
0.8
1.0
36.2
22.3
17.0
14.1
12.7
62.7
47.7
38.9
33.1
29.1
92.6
83.9
75.6
68.2
62.1
Notes:
• Calculations are for a purely redistributive tax: i.e. R = 0
• Broome case corresponds to bottom right corner. But he
assumed that there was no-one below 70.71% of the median.
Frank Cowell:
“Linear” model: assessment
UB Public Economics
Solution to problem becomes much more transparent
But exact tax formulas are still elusive.
Optimal tax rates are very sensitive to precise assumptions
about
labour-supply elasticity.
Distribution of ability
Inequality aversion
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
An alternative
focus on human
capital
General labour
model
“linear” labour
model
Education
model
Generalisations
Frank Cowell:
Approach 3: Alternative Income
Determination
UB Public Economics
Most OIT models focus on just one area of personal
decision making
Casual discussion of policy suggest that other economic
incentives may be relevant
What about the long-run determination of earning power?
Need a model of investment.
Frank Cowell:
Components of Atkinson’s
human capital model
UB Public Economics
Given structure of ability distribution
Individuals maximise lifetime disposable income
Essentially investment model
Based on Becker (and Mincer) human capital model
Schooling only, not experience
Conventional social welfare function
Government budget constraint of zero net revenue
Frank Cowell:
Notation in Atkinson’s human
capital model
UB Public Economics
w - exogenously given ability
y - pretax income
S - years of schooling
L - length of working life
r - interest (discount) rate
c - disposable income
UB Public Economics
earnings
Frank Cowell:
Life Cycle in the Atkinson Model
y=wS
t
S
L+S
age
Frank Cowell:
Atkinson’s Becker-typePareto
approach
distribution of
ability
UB Public Economics
pretax income determined
by Becker schooling model
choose schooling to
maximise discounted
lifetime consumption
Frank Cowell:
Atkinson’s human capital model:
optimised schooling
UB Public Economics
Disposable income is c = B + [1-t] y
Define a critical ability level in terms of tax
parameters
Ability type w chooses optimal schooling as
For medium/high ability schooling increases with ability
For low ability it’s not worth investing in education
Frank Cowell:
Atkinson’s human capital model:
optimised utility
UB Public Economics
Substitute optimal S into formula for
discounted lifetime consumption to get:
Gives relationship between ability and
utility
Frank Cowell:
Atkinson model: social
objectives and constraints
UB Public Economics
Maximise additively separable SWF as
before.
Government budget constraint becomes
Frank Cowell:
Atkinson’s human-capital model:
income will become
pretax and disposabletaxable
income
more unequal the more
progressive is the tax
UB Public Economics
disposable income will have the
same inequality as ability!
Frank Cowell:
Ability-Schooling Relationship
for values of w0 = rB/[1 – t]
0.12
schooling
UB Public Economics
0.1
0.08
0 (No tax)
0.05 (low prog)
0.10 (medium)
0.15 (high)
0.06
0.04
0.02
0
150
200
250
300
350
400
ability
In a high-progression model the able invest a lot in education
This pays for the income supplements for the less able
Frank Cowell:
Atkinson’s “Becker” model:
optimal marginal tax rates
UB Public Economics
Frank Cowell:
Education model: assessment
UB Public Economics
Key to model is investment response to anticipated tax
In simple model schooling chosen increases when tax
progression is increased.
Result can appear to offset effect on current income
But target is distribution of lifetime utility.
Result of low optimal marginal rates depends crucially on
appropriateness of the precise investment model
Frank Cowell:
Overview...
Optimal Income
Taxation
Design
Issues
UB Public Economics
What if we
combine insights
from the two
main branches of
optimal taxation?
General labour
model
“linear” labour
model
Education
model
Generalisations
Frank Cowell:
More general tax issues
UB Public Economics
Should we rely on direct or indirect taxation?
Is there much to be gained by combining the two
branches of theory?
Can a unified optimising model be developed?
Frank Cowell:
Direct versus Indirect Taxation
Issues
UB Public Economics
1.
Nonlinear commodity taxation?
2.
Informational requirements.
3.
Participation and incentive compatibility
constraints.
4.
Direct versus indirect tax progressivity.
Frank Cowell:
1 Nonlinear commodity taxation?
UB Public Economics
Should consider the issue of proportional versus
nonlinear taxation of commodities.
“Nonlinear” includes affine functions (like the socalled linear income tax function).
The argument is whether each commodity should
be “repriced”, perhaps not in a proportional
fashion.
Similar argument is applied in other areas: tariffs
for output of state-owned industries, price support
schemes
Frank Cowell:
2 Informational requirements
UB Public Economics
Recall the main differences between the two types of tax:
Not the formal tax base (income versus expenditure) but the
informational base.
Direct tax authority can know details of personal resources.
Indirect tax authority can know structure of production and
transactions
Informational requirements may preclude extensive
application of nonlinear commodity taxes.
To see this consider problem of nonlinear pricing of
consumer goods.
Can work for water, gas, electricity
But for food? Clothes?
Frank Cowell:
3 Participation and Incentive
Compatibility Constraints
UB Public Economics
ICC issues are central to nonlinear income tax design
Same difficulty can arise with nonlinear pricing schemes:
Some groups may choose the “wrong contract”
Arises both in private and public sector
Difficulties usually disappear if you impose the regularity
conditions implied by linearity
Supports the strong case for considering linear commodity
taxes
Frank Cowell:
4 Direct versus Indirect Tax
Progressivity
UB Public Economics
Can measure progressivity in a number of ways
A standard method is to compute the implied tax rates that
emerge from actual expenditure decisions
Can do this for the definitions of “direct” and “indirect”
taxes in the UK
In practice indirect taxes are more regressive than direct
taxes.
Frank Cowell:
Implied average tax rates in
Economic Trends. UK 1994
0.35
UB Public Economics
0.3
Direct
Indirect
0.25
0.2
0.15
0.1
0.05
0
Bottom
10th
2nd
3nd
4th
5th
6th
7th
8th
9th
Top
10th
Frank Cowell:
Integrating direct and indirect
taxation: consumer’s problem
Total disposable income is given by
UB Public Economics
.
so the budget constraint is:
Assume there is no lump sum income (I=0)
Frank Cowell:
Integrating direct and indirect
taxation: government’s problem
UB Public Economics
Government budget
constraint is
otherwise you’ll get lump
sum taxation again!
.
Given the generality of the problem we should
reduce the number of degrees of freedom
Use this
to give general guidance on tax
First order conditions
yield
structure.
Frank Cowell:
Policy rules
UB Public Economics
Commodity taxes should be zero if preferences are
weakly separable in leisure and other goods
Tax on good i should be higher if the MRS between
good i and labour increases.
Focus tax on goods for which the most able have the
strongest preference.
Frank Cowell:
Conclusions
UB Public Economics
Direct versus indirect
Distinction between the two is essentially an issue
of information.
Big differences in terms of distributional effect.
Uniform commodity taxation
No compelling case within the context of the model
There may be a case if you appeal to other factors
“Flat tax”
Argument as for uniform commodity taxation