No Slide Title

Download Report

Transcript No Slide Title

Sydney February 2007
The Evolution of the
Distribution of Individual
Earnings in OECD Countries
A B Atkinson, Nuffield College, Oxford and
Paris School of Economics
1
Introduction
1. Taking the Supply and Demand Story Seriously
•
Modelling the race between technology and education
•
Quantities and Prices
2. Empirical Evidence about Earnings Dispersion
•
An augmented dataset on earnings for 18 OECD countries
•
Patterns of change
3. Enriching the Story
•
A Behavioural Model of Changing Pay Norms
•
Earnings at the top: superstars and managerial pyramids
Conclusions
2
3
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
Decile ratio P90/P10
Increasing decile ratio in the US 1980-
4.8
4.6
4.4
4.2
4.0
3.8
3.6
3.4
3.2
3.0
2.8
Were the 1990s a Pause
or
have we reached a Plateau?
or ?
4
Textbook Story
There is general agreement that the main factor behind
the increase in the relative wage of skilled versus
unskilled workers is a steady increase in the relative
demand for skilled workers. (Blanchard,
Macroeconomics, page 524).
Globalisation: fall in relative price of goods in which
unskilled workers are more intensively engaged.
Technical progress biased in favour of skilled workers.
5
Relative wage of qualified
workers
ws / wu
denoted by ω
●
●
Shift in Demand
Relative demand for
qualified workers Ls/Lu
6
Missing
• Dynamic story but dynamics not modelled.
• Distribution of earnings depends on quantities
as well as prices; cannot just look at wage
differential.
• “New” theories of labour market – Section 3.
• Inter-relation between labour and capital
markets.
K Arrow and W M Capron, “Dynamic Shortages”,
QJE 1958
7
Simple temporary equilibrium model
At any point in time, supply of qualified workers is fixed at S and
relative wage ω clears the market
D[ω] = S
Over time, demand shifts at exogenous rate g
dlnD/dt = g – σ dlnω/dt
(1)
Supply adjusts according to premium over cost of acquiring
education, which is assumed to be simply that of postponing
entry to paid work by T years, r is the cost of borrowing
dlnS/dt = β [ ω - e rT ]
(2)
Hence
dlnω/dt = (β/σ) [g/β + e rT - ω]
8
dlnω/dt = (β/σ) [g/β + e rT - ω]
β zero means pure rent, differential grows without limit
β positive means convergence
β infinite means pure compensating differential e rT
Relative
earnings
g/β + e rT
erT
9
Relative wage of qualified workers
ws / wu
●
●
+ g/β
erT
Relative demand for qualified
workers Ls/Lu
10
• Casts doubt on criticism of the Skill Biased Technical
Change (SBTC) hypothesis: “our main conclusion is that
aggregate wage inequality has not risen continuously since
the 1970s. … This suggests a potential problem for the
SBTC hypothesis: why did the pace of SBTC slow down
after an initial burst during the first few years of the
microcomputer revolution?” (Card and DiNardo, JLE, 2002).
• But Also: Wage premium depends on β, which may vary
from country to country.
• Wage differential may have increased because of rise in the
cost of borrowing
= inter-connection between labour and capital markets.
11
HOWEVER Distribution depends on
Changes in Quantities as well as on
Wage Differential
• Kuznets model = model of quantity changes.
• Societies becoming more educated = same kind of
structural change.
• (With constant wage premium) tends to shift Lorenz
curve outwards at bottom and inwards at top.
• ? Increase in dispersion among qualified workers ?
12
Lorenz Curve
Share of
total
earnings
Increase in
dispersion
among the
skilled
Effect of rise in ratio
of skilled to unskilled
in labour force
SKILLED
UNSKILLED
P10
P 90
Share of total population
13
2. Empirical Evidence about Earnings
Dispersion
OECD View
“ Gross earnings inequality has increased on average
in OECD countries for which data are available. This
occurred in countries where labour market
performance improved considerably (Australia, the
Netherlands), as well as in countries where it
deteriorated (Germany, Hungary, Korea and Poland).
Broadly unchanged wage dispersion was recorded
in [France, Japan, Switzerland and UK]”.
Assessing the Jobs Strategy May 2005
14
OECD comparison of 2000 and 1994
2000 ratio
5
HUN
USA
4
KOR
P OL
UK
DEU
A US
FRA
3
NLD
JP N
CHE
FIN
SWE
2
2
3
4
5
1994 ratio
15
Decile ratio
• Short time period
• Series are not necessarily consistent over
time
• Looks at decile ratio, not bottom and top
deciles separately
• No clear metric: what constitutes a “rise”?
16
An Augmented Database
•
18 OECD countries
• Identifies breaks in series
• Multiple sources for most countries
17
Gottschalk
and
Smeeding
(1997)
AUS,
CA, FI,
D, ISR,
NL,
SWE,
UK, US
“Almost all industrial economies experienced some increase
in wage inequality among prime-age males during the 1980s
(Germany and Italy are the exceptions). … But large
differences in trends also exist across countries, with earnings
inequality increasing most in the United States and the United
Kingdom and least in the Nordic countries”
Katz and
Autor
(1999)
AUS,
OS,
CA, FI,
FR, D,
IT, JA,
NL, NZ,
NOR,
SWE,
UK, US
“The United States and the United Kingdom experienced
sharp increases in overall wage inequality … The pattern of
declining wage inequality apparent throughout the OECD
(except the United States) in the 1970s ceased in the 1980s
and 1990s in almost all nations (with Germany and Norway as
possible exceptions). Canada, Australia, Japan, and Sweden
had modest increases in wage inequality … Wage inequality
narrowed through the mid-1980s in Italy and France with
some hint of expanding in France in the late 1980s and with a
large increase in inequality in Italy in the 1990s … New
Zealand also shows large increases in inequality”
18
• 1979/1980/1981
• versus
• 1999/2000/2001
Metric
Five percent criterion for change to “count”
P10 50
P90
180
DR
3.60
P10 47.5
P90
189
DR
3.98
Significant = increase of Ten percent or more
19
ria
ly
N
S
K
U
U
nd
Po
rt
ug
a
Sw l
ed
en
Z
L
or
w
ay
N
Ita
Po
la
N
D
R
en
m
ar
k
Fi
nl
an
d
Fr
an
ce
G
er
m
an
y
H
un
ga
ry
ze
ch
da
us
t
ra
lia
an
a
A
us
t
C
-5
C
A
Change in Bottom Decile % 1980-2000
15
10
5
0
-10
-15
20
or
w
L
ay
N
S
K
U
U
N
Po Z
la
Po nd
rt
ug
Sw a l
ed
en
N
Ita
ly
us
tr
al
A ia
us
tr
ia
C
an
a
C da
ze
c
D hR
en
m
a
Fi r k
nl
an
Fr d
a
G nce
er
m
an
H
un y
ga
ry
A
Rise in Top Decile % 1980-2000
25
20
15
10
5
0
21
Earnings deciles in the UK (1977 = 100)
1.20
Fanning Out
1.15
P80
Top decile
1.10
P70
1.05
P60
STABILITY
1.00
0.95
0.90
Fall
Bottom decile
Some recovery
P40
P20
and
P30
0.85
0.80
1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005
22
Australia Results from Employee Earnings and Hours Survey
200
Top decile
180
160
140
120
100
80
60
40
Bottom decile
20
0
1974
1979
All workers ■
1984
1989
FT adults ‫٭‬
WEE (for comparison) ∆
1994
1999
2004
FT non-managerial ○
23
Share of top 1% in total EARNINGS
11
US
% share total earnings
10
9
Share of Top 1%
FRANCE
8
CA
7
6
5
UK
4
3
1946 1950 1954 1958 1962 1966 1970 1974 1978 1982 1986 1990 1994 1998
24
Summary of empirical evidence
• Important to look separately at bottom and top
deciles
• 7 out of 18 saw fall in bottom decile of 5% or more
from 1980 to 2000
• Majority ( 14 out of 18) saw rise in top decile; in 7 out
of 18 rise exceeded 10%
• In UK and US, decile ratio broadly stable since 1990
because rise in bottom decile offsets rise in top
decile
• Behaviour at top differs across countries.
• Need to explain distribution among the qualified.
25
Enriching the Story
Two of many possibilities:
• Behavioural pay model, with shift in
pay norms (Atkinson, 1999)
• Superstar rents and managerial
pyramids at top
26
Employer Conform
Not conform
Worker
Conform
Match
No match
Not conform
Pay follows
code
No match
Match
Paid
according to
individual
productivity 27
100%
E3
Jobs
adhering
to pay
norm
E2
E1
0
100%
Belief about degree of acceptance of pay norm
28
100%
E3
Jobs
adhering
to pay
norm
E2
E1
0
E1*
Firms with higher
discount rates attach
less weight to
reputation
100%
Belief about degree of acceptance of pay norm
29
Superstar Theory
(Alfred Marshall 1890s and Sherwin Rosen 1980s)
+ Gives role to both technology and trade
- No direct link to distribution
? Explain earlier periods when top earnings fell?
30
Log (Earnings/median)
Superstar model
generates extreme value
distribution with Pareto
tail with exponent α
Effect of trade and
technology in
expanding share of
rents captured by top
performers = fall in α
Slope = 1/α
Log [1/(1 – F)]
31
Managerial Hierarchy Model (Lydall and
Simon)
7.2
β=
loge[span of managerial control]
divided by
loge[1+ increment with promotion ]
increment
span
5
25%
32
Log (Earnings/median)
Superstar model not enough on
its own, since not explain earlier
rise in α
+
Hierarchical Salary Model
Hierarchical model not enough
on its own, since predicted
Pareto exponent β too large
Log [1/(1 – F)]
33
Not just a simple Race between
Technology and Education
• we have to study dynamics; there may be a temporary
equilibrium with constant premium for qualified workers;
there may be cobweb cycles.
• Distribution depends on quantities as well as wages.
• There has been widening of earnings dispersion in majority of
OECD countries, but rises in decile ratio due more to
changes at top than at bottom.
• Need to take account of “new” theories of the labour
market, such as behavioural models of pay norms.
• Top earnings could be explained by mixture of hierarchy
and superstar theories.
34