j.roemer "economic development as opportunity equalization"
TRANSCRIPT
““ Economic development as Economic development as
equalization of equalization of
opportunitiesopportunities ””
John E. Roemer
Yale University
1
How to measure economic
development?
Classically, as GDP per capita
The normative justification:
Individual Welfare = income yi Social welfare is the sum or average of incomes: 1N
yi∑
Can be criticized in at least two ways:
o welfare is linear in income, which ignores the
urgency of some needs over others
o social welfare is utilitarian, and reflects no
concern for inequality
2
• A technological justification?
You might say: GDP per capita is not intended as a welfare
measure but as a measure of industrial capacity.
But economic development must mean the advance of
human
society; it cannot be a technological concept. A productive
technologyrun by slaves all of whose product goes to a
small elite shouldnot be considered a highly developed
economy.
So I insist the justification of an index of econ development
be corollary to a concept of social welfare
3
Various possibilities
4
Equality of opportunity
The distribution of the social objective (here
income) should be independent of
circumstances beyond the control of
individuals, but may reflect actions that are
within their control
‘Leveling the playing field’ metaphor means
compensating persons for the effect on their
achievments which reflect disadvantages
beyond their control. The ‘troughs’ in the
playing field are these disavantages
5
Language
• Circumstances are those aspects of a
person’s environment that influence
outcomes & are beyond his control
• The typology is the partition of the pop’n
into types, where all individuals of a type
have similar circumstances
• With a type, outcomes will differ because
of differential effort
6
Example: The distribution of income
by type, defined as level of
parent’s education in Austria
(2005)
7
Contrast with Denmark and Hungary:
8
Define ut (e) as the outcome for an individual of type t who
expends effort e. ut is monotone increasing in effort: so it is not a
classical utility function
Let the distribuition of the outcome in type at the status quo be
Ft (u). How can we compare the efforts of two individuals in
the same type? Easy.
In different types? Difficult. Observe that the distribution of effort
is itself a circumstance. So to properly define degree of effort, must
sterilize this distribution of the effect of circumstances upon it. I
propose: Measure the degree of effort of an individual by her rank
on the effort distribution of her type; but this is also her rank on the
outcome distribution of her type!
Thus the EOp prescription is: Design policy to minimize the
horizontal distances between the CDFs of the objectives of the types
9
Formally:
Define as the objective value of the folks in
type t at quantile of the distribution of the objective
when the policy is
Choose policy ϕ to
maximize .
If the policy is the status quo, I define:
.
This is the mean value of the objective for the most
disadvantaged type. It is the area to the left of the most
disadvantaged type’s CDF and bounded above by the line
at 1.
vt (π;ϕ )πϕ
mint
0
1
∫ vt (π;ϕ ) dπ
W EOp = mint
0
1
∫ vt (π;ϕ s.q. ) dπ
10
So the EOp prescription is : design policy to push the CDFs
to the right as far as possible.
11
To what extent is unequal
opportunity responsible for
inequality? Let C(F) be the coefficient of variation
squared of a distribution F. If we have a
partition of the pop’n into types, and F is the
aggregate distribution, we can decompose
C(F) into a sum of two pos numbers:
C(F) = C(ΦT )+ (ρ t∑ )2 ftC(F
t )
12
where ΦT is the counterfactual distribution of
income where every member of type t earns
exactly the average income of type t.
13
It’s therefore natural to think of C(ΦT )C(F)
as
a lower bound on the fraction of total
inequality due to circumstances. Therefore,
an upper bound on the degree of inequality
due to effort is:
ηupp bnd = 1− C(ΦT )
C(F).
14
We can also perform a ‘dual
decomposition.’ Partition the society into
tranches π where each tranche comprises the
set of individuals (of all types) who expended
effort of degree π . We have a
decomposition: C(F) = C(E)+ (ρπ )2∫ C(Eπ )dπ
where E is a hypothetical distribution in
which everyone who expended the same
degree of effort has the same income, and Eπ
is the distribution of income in tranche π .
In like manner, we can interpret
ηlwr bnd = C(E)C(F)
as a lower bound on the
degree to which effort accounts for inequality
15
I therefore propose as a measure of the
degree to which effort explains inequality as
η =12(ηupp bnd +ηlwr bnd ).
16
A two-dimensional index of
economic development (Wc
EOp ,ηc )
The first component measures how well the
most disadvantaged type is doing and the
second measures the degree to which effort
(as opposed to circumstances) explains total
inequality in the society
17
Application
Use the data set EU-SILC to compute the
index (W EOp ,η) for a set of 22 countries.
• Circumstance: Typology into three types
based upon level of parental education
• Objective: Income after taxes & transfers
in 2005. (note defect: does not include
the value of public goods)
18
AT
BE
CZ
DK
DE
EE
ES
FI
FR
GR
HU
IS
IT
LT
LU
LV
NL
PL
PT
SESI
UK
10000 20000 30000 40000level
0.80
0.85
0.95
degree
Levels & Degrees of development
19
Some comments:
0. Levels of η are high
1. Lowest levels W EOp are the east
European countries & Spain
2. The only undominated country is
Denmark
3. Iceland(IS) looks good: but this is pre-
crash
4. I don’t believe Greece’s value of η .
20
Country, UBE LBE average effort
AT 0.970434 0.92 0.945217
BE 0.982783 0.646 0.814392
CZ 0.919772 0.892 0.905886
DK 0.990947 0.976 0.983473
DE 0.995659 0.746 0.870829
EE 0.956504 0.831 0.893752
ES 0.953383 0.883 0.918191
FI 0.988751 0.954 0.971375
FR 0.971633 0.886 0.928817
GR 0.994 0.945 0.9695
HU 0.946974 0.604 0.775487
IS 0.984799 0.923 0.953899
IT 0.985046 0.777 0.881023
LT 0.965407 0.92 0.942704
LU 0.964373 0.747 0.855687
LV 0.964469 0.723 0.843734
NL 0.991502 0.878 0.934751
PL 0.983872 0.781 0.882436
PT 0.95876 0.834 0.89638
SE 0.984744 0.93 0.957372
SI 0.982 0.944 0.963
UK 0.991047 0.703 0.847023
21
All highly developed
countries
In LDCs, both indices (W EOp ,η) will be
much smaller. For Latin America, η will be
typically under 0.5.
22
Question: Is there value added
using this measure vs. GDP per
capita?
Let’s aggregate the two dims of development
as (W EOp )θη1−θ for various choices of
θ ∈[0,1] and ask: What’s the rank-
correlation between the numbers (W EOp )θη1−θ
by country and the order of GDP per capita
by country?
Answer: For θ > 0.3, Spearman coefficient
of rank correlation is over 0.95! Of course
we lose information by passing to a uni-dim
measure
23
Why? Two reasons:
(a) η > 0.80 for all countries in the
sample
(b) correlation of the order of WcEOp and
GDPc is very high
But in a panel of countries of the world, we
would see some more dramatic differences
between the two rankings – in the main
because values of η will be much smaller
countries at low levels of development
24
Economic development is equity
World Development Report 2006 : Equity
and Development . An excellent report.
But suffers confusion of implicitly assuming
that development is measured by GDP per
cap, and counterposing this with ‘equity’
conceived of as EOp
25
Some quotations from WDR 2006:
“Greater equity is thus doubly good for poverty reduction: through potential
beneficial effects on aggregate long-run development and through greater
opportunities for poorer groups within any society (p.2)”
“If the opportunities faced by children like N. are so much more limited than those
faced by children like P. or S., and if this hurts development progress in the aggregate,
then public action has a legitimate role in seeking to broaden opportunities….(p.3)”
“Third, the dichotomy between policies for growth and policies specifically aimed at
equity is false (p.10)”
26
Further discussion
• Choice of income as the objective and typology based
on parental education is suggested to enable us to
calculate (W EOp ,η) for a very broad set of countries
• Ideally include more circumstances: rural vs urban,
race/ethnicity, gender
• For Sweden, Bjorklund, Jantti and Roemer perform a
typology of 1152 types using a huge Swedish data set.
We compute (using a different method) that
circumstances are responsible for 33% of inequality –
as opposed to the measure here of ηSweden = 4.3%.
• Even in egalitarian Sweden, circumstances remain
important in the determination of income.
27
Treatment of children • Children are not responsible until an ‘age of consent’
–perhaps age 15. Nature and nurture are both
circumstances in the discussion of child achievement.
• So if we want to look at incomes at age 30 (say), we
should treat measures of accomplishment at age 15
(education, literacy, etc.) as circumstances. Not to do
so is to hold children responsible for what they
achieve below the age of responsible action.
• See World Bank, Measuring inequality of opportunity
in Latin America and the Caribbean for an excellent
analysis of that topic, and calculation of inequality of
opportunities with regard to child accomplishment
28
Are people responsible for
their preferences?
• No: not entirely. The child who grows up in a
home with parents who have not experienced
education should not be responsible for having a
preference against higher education
• To say such preferences are due only to ‘lack of
information’ is a cop-out
• My proposal compensates persons for the effect
of circumstances on the distribution of effort
choices in their type. Since, neo-classically,
effort choices result from preference-satisfaction
maximization, we do not hold persons
responsible entirely for their preferences, but
attempt to sterilize away the effect of
circumstances on choices. What variation in
29
choice remains after this sterilization is held to be
the person’s responsibility
• This is to some degree paternalistic. It is a liberal
fallacy that paternalism is an unmitigated evil
30
EOp versus Meritocracy • In a meritocracy, individuals receive incomes which
are monotonically related to their skill levels. In an
ideal EOp distribution, they receive incomes
montonically related to their effort. These are
different rules. A meritocracy rewards skills even if
they are due to circumstances, while EOp in principle
does not.
• Since markets rewards skills independently of their
cause (circumstance or effort), markets are meritocrats
are more friendly to markets than opportunity
egalitarians.
• E.g., In my view, a person should not be materially
deprived because he lacks native intelligence. But
implementing that kind of talent-blind income
distribution requires strong market intervention via
transfers, etc.
31
Scope
• But I do not propose including IQ as a circumstance in
the calculation of the indices of economic
development today. I propose limiting circumstances
to measures of socio-economic advantage (like
parent’s education & income), race, etc. Including
IQ as a circumstance is an ethical advance appropriate
for a future time
• A friendly amendment to my proposal is to meld the
human development index and EOp: Take the
measure of advantage as the HDI, but disaggregate
the distribution of individual human development by
type, and proceed as I have in this paper.
32
EOp is good policy • Equality of opportunity is an ethic endorsed by
billions of people. In many surveys, respondents
answer that the justice of the income dist’n depends on
the extent to which incomes are due to effort versus
luck. This is the distinction EOp captures.
• The utilitarian measure of GDP per cap is insensitive
to this distinction
• The HDI as it is currently measured is insensitive to
this distinction – even with its latest inequality-
sensitive variant
• Computing the two–dimensional index of economic
development described here is consonant with popular
views of justice, and will lead to different policy
proposals than the measure of GDP per capita.