Narrowing, Widening, Polarizing: The Evolution of the U'S' Wage Structure

1
Narrowing, Widening, Polarizing The Evolution of
the U.S. Wage Structure

• Lawrence F. Katz, Harvard University
• With special thanks to Claudia Goldin, David
  Autor, Kevin M. Murphy, and Richard Freeman for
  collaboration and guidance

2
Research on the Wage Structure Has a Long
Tradition in Economics
Douglas
Smith
Tinbergen
3
Outline for Talk

• 1. Recent U.S. Wage Structure Changes
• 2. Why SBTC?
• 3. Something Standard (Widening) Supply-Demand
  Analysis of Changes in College Wage Premium,
  1963-2003
• 4. Something Old (Narrowing) The Decline of
  Non-Competing Groups, 1890-1950
• 5. Something New (Polarizing) The Polarization
  of the U.S. Labor Market, 1988 -- ???

4
Theme Race between Technology and Education

• (1) Rapid Secular Growth in Demand for Skills
• (2) Variation in rate of growth of supply of
  skills is key factor acceleration around 1910
  and deceleration post-1980
• (3) Impact of computers on labor market is more
  subtle than standard monotonic SBTC view manual
  vs. routine vs. abstract tasks

5
Rising U.S. Wage Inequality since 1980

• Major patterns Full-Time Weekly Wages
• Rising educational differentials. College Plus/HS
  gap
• 40 log points in 1980 ? 63 log points in 2003
• Rising overall inequality. 90/10 log wage ratio
  (Males)
• 126 log points in 1980 ? 163 log points in 2003
• Rising residual inequality. 90/10 log residual
  ratio (Males)
• 105 log points in 1980 ? 127 log points in 2003
• Large lit concludes and much evidence indicates
• A secular, long-term phenomenon over 2 decades.
• Large role shifts in the supply of and demand for
  skills
• Eroding institutions affecting low-
  middle-wage workers
• Source Autor, Katz, and Kearney (2005, NBER WP
  11627)

6
Change in Log Real Weekly Wage by Percentile,
1963-2003
7
Measures of FT Weekly Wage Inequality, March CPS
8
Measures of Hourly Wage Inequality, 1973-2003
9
Male 90/50 Wage Inequality 1963 - 2003
10
Male 50/10 Wage Inequality 1963 - 2003
11
Why SBTC?

• INDIRECT EVIDENCE
• Within-Sector and Within-Establishment Skill
  Upgrading in Face of Rising Skill Prices
  (Education, Occupations, Task Mix)
• Cross-Country Similarity of Skill Upgrading
  Patterns (Berman, Bound and Machin 1998)
• Changes in P (wages) and Q (employment)
  correlated across groups (especially education)
  groups implying demand shifts
• But Between Industry (and Between-Plant) Demand
  Shifts Not Large Enough to Explain Observed
  Patterns

12
Evidence of Demand Shifts ?
13
Why SBTC?

• DIRECT EVIDENCE
• Technology-Skill and Capital-Skill
  Complementarity Positive Correlations of
  Technology Indicators (Computers, RD, New
  Equipment) and Skill (Education, Occupations,
  Tasks) of Work Force in Levels, Changes and
  Acceleration (by Industries and Establishments)
• Many Compelling Case Studies of Technology (IT)
  Investments, Workplace Re-Organization, and Skill
  Mix of Work Force for financial services, auto
  repair, valve manufacturing, steel, etc. (Autor,
  Murnane and Levy Batel, Ichniowski, and Shaw)

14
Secular SBTC vs. Acceleration Hypothesis

• Similar cross-industry and cross-firm patterns of
  technology-skill (re-organization of production
  with shift to electricity) and capital-skill
  complementarity for 1890-1929 (Goldin-Katz 1998)
• The indirect and direct evidence strongly suggest
  secular SBTC increasing demand for more-educated
  workers at least since 1890
• Has there been a demand acceleration since 1980?
• Alternative View Recent period differs from
  past because of slowdown in growth of supply of
  skills more than from an increase in growth in
  demand for skills

15
Supply-Demand Model for College Wage Premium

• The Standard Explanation Rising demand for
  skill. Katz-Murphy (1992), Murphy-Welch (1992),
  Bound-Johnson (1992), Autor-Katz-Krueger (1998),
  Card-Lemieux (2001).
• Given information on Wct/Wht, Nct/Nht, and an
  estimate of s...
• Can infer relative demand shifts for college vs.
  high-school labor.
• Secular demand shifts for college workers
  combined with supply fluctuations works
  reasonably well.

16
Log Relative Supply of College/Non-College Labor,
1963 - 2003
17
Slowdown in Growth of Educational Attainment
18
(No Transcript)
19
Estimation of relative demand shifts

• Simple Implementation (Katz-Murphy, 1992)
• ln(Wct/Wht) g0 g1?t g2 ? ln(Nct/Nht) et
• Where g1 estimates the trend on the demand index
  (1-1/s)?ln(Act/Aht)
• g2 is an estimate of 1/s, aggregate elasticity of
  substitution.
• CES Production function with college and HS
  equivalents
• Katz-Murphy estimated that g2 ? .71 ? s ? 1.41
• Can also
• Use alternative estimates of s
• Augment estimation with other covariates.
• Two-level CES with imperfect substitution among
  different experience groups use cross-cohort
  variation in college relative supply and size

20
(No Transcript)
21
(No Transcript)
22
Growth of College/HS Relative Wage, Supply and
Demand, 1940-2005 (100 x annual log changes,
sigma 1.4)
23
Decline of Non-Competing Groups, 1890-1950

• Technological Revolution also occurred early in
  20th Century
• Electrification of the Factory Mfg Horsepower
  from Purchased Electricity up from 9 in 1909 to
  53 in 1929 declining electricity price similar
  to todays declining computer prices
• New Goods and Production Processes Proliferated
  autos, airplanes, aluminum, radio, synthetic
  dyes, rayon, office machinery, household
  appliances
• Huge increase in complexity, scale of enterprise,
  scale of markets and demand for white collar
  (clerical, managerial, prof) workers
• SBTC Strong cross-section evidence of
  technology-skill and capital-skill
  complementarity
• But the best evidence indicates narrowing
  inequality and skill differentials in 1st half of
  20th Century even before 1940s

24
Narrowing Wage Inequality, 1890-1950

• Occupational Wage Ratios available prior to 1940
  Census
• -- Skilled (Machinists, Printing Compositors,
  Building Trades) relative to Less-Skilled Manual
  Trades
• -- White-Collar (Clerks, Professors) to
  Blue-Collar (Mfg production worker) Wage
• Within-Industry Wage Distributions for Mfg, 1890
  to 1940s
• Direct Evidence of Educational Wage Differentials
  from
• Iowa State Census of 1915
• -- 1914 Occupational Earnings and Years of
  Schooling by Type of Schooling (Common, Grammar,
  HS, College)
• -- Compare Iowa in 1914 to 1940 and 1950 Censuses
• Source Goldin and Katz (1995, 1999, 2000, 2001)

25
(No Transcript)
26
Returns to Schooling Iowa State Census, 1915
27
Educational Wage Differentials, 1915-2000
28
Why Did Skill Differentials Decline Prior to 1950?

• Rapid SBTC and Growth in Demand for Skilled (HS)
  Workers
• Cobb-Douglas White Collar Demand Growth Index in
  Mfg
• Annual change in log(nonproduction/production
  worker wage bill ratio)
• 1890-1929 1.03
• 1959-1989 1.27
• 1965-2004 1.04
• Key Roles of World Wars I and II -- Timing
  Issues
• Rising Supply of Skills not Falling Demand is Key

29
Why Decline of Education Returns and Skill
Differentials in First-Half of 20th Century?

• High School Movement Takes Off circa 1910
• -- Rapid and Accelerating Growth in Relative
  Supply of Skills
• -- High returns to HS with growth in demand for
  office workers and skilled production workers
• -- Lower individual costs with building of
  public high schools
• -- State free tuition laws
• -- Expands first in wealthy agricultural areas,
  small towns
• -- Key role of local control, egalitarianism,
  social capital
• -- Lower opportunity costs with decline of child
  labor intensive sectors, further decline with
  high unemployment in 1930s
• See Goldin (1998) and Goldin and Katz (1997,
  1999, 2003, forthcoming)

30
OGDEN HIGH SCHOOL
31
(No Transcript)
32
Public and Private Secondary School Enrollment
and Graduation Rates
33
HS Graduation Share by Birth Cohort, IPUMS
34
Educational Growth Accounting, 1915-2000

• Change in Change in Mean
• Educational Years of Schooling
• Productivity of Work Force
• (annualized)
• 1915-40 0.52 1.38
• 1940-60 0.50 1.52
• 1960-80 0.61 1.93
• 1980-2000 0.35 0.86
• Chain-Weighted Prices
• Source DeLong, Goldin and Katz (2003)

35
Something New Polarization of Labor Market,
1989 - ???

• A polarization of employment and earnings over
  last 15 years (Autor, Katz, Kearney 2006 NBER WP
  11986 see Goos and Manning for UK)
• Definition Polarization (Merriam-Webster)
• Division into two opposites.
• Concentration about opposing extremes of groups
  or interests formerly ranged on a continuum.
• Convexification of returns to education over last
  15 years (Lemeiux 2006).
• Mean Percentile in Wage Distribution (2000 CPS
  MORG)
• HS Graduate 40
• BA Only 66
• Graduate Degree 77

36
The Divergence of Upper- and Lower-Half Inequality
37
Real Wage Growth by Wage Percentile Monotone
1979 1987, U-Shape 1987 Present
38
Polarization Hypothesis

• Exogenous Driving Forces Computerization
  (Declining Computer Prices) and Globalization
  (Reduced Transport/Communication Costs China,
  India)
• ALM Production Function with Routine and
  Non-Routine Tasks Price of Routine Tasks
  Declines Substitute Computers and Offshoring
• Reduced Demand for Routine Manual Jobs (Blue
  Collar) and Routine Cognitive Jobs (Insurance
  Processing, Programming, Call Centers) --
  declining middle
• Increased Demand for Problem Solving -- Larger
  Teams for High-Level Managers and Professionals
• Little impact on non-routine manual, in-person
  services

39
Linking Polarization with Market Forces

• Consider the jobs in demand 1980/1990/2000
• Several approaches
• Measuring employment growth of by occupational
  skill
• Skill defined by occupational mean education
• Skill defined by occupational wage levels
• Measuring changes in job tasks
• Tasks defined using Dictionary of Occupational
  titles (as in ALM 2003)
• Do we see?
• Monotone employment growth by skill in the 1980s?
• U-Shaped employment growth by skill in the 1990s?

40
Changes in Occupation Employment Shares
1980-1990, 1990-2000Occupations Ranked by
Average Years Schooling 1980
41
Polarization of Employment Found in Other
Countries
(1) UK Job Growth by Quality 1979-1999
(2) W. Germany Job Growth by Skill 1979-1999

• Goos and Manning 2003 (Left-Hand Figure)
• Polarization of employment in UK over 1979
  1999
• Spitz-Oener 2006 (Right-Hand Figure)
• Polarization of job tasks in western Germany over
  1979 1999

42
Alternative Approach Job Task Content

• Autor, Levy, Murnane 2003
• Conceptualize work as being made up of sets of
  tasks
• Examine how job tasks have evolved over multiple
  decades
• Method
• Link Dictionary of Occupational Titles to
  occupations in 1973/79.
• Use changes in occupation distribution over 1959
  2002 to measure changes in job task content.
• Key observations of ALM approach
• Computer capital has comparative advantage in
  Routine tasks.
• Human labor has comparative advantage in tasks
  requiring Problem Solving (Abstract) or
  Flexibility (Manual).
• Real price of computer power falls precipitously
  Nordhaus 2001
• One-trillion fold since 1900. 55 percent per year
  since 1940.
• May change what tasks workers do and the wages
  paid to those tasks.

43
Representative EvidenceTrends in Job Task
Content 1960 2002
44
Why Task Trends Foster PolarizationTask Input
by Major Education Group in 1980
45
A Simple Model of Computerization and Polarization

• Three categories of job tasks
• Routine (R) Bookkeeping, clerical work,
  repetitive production
• Abstract (A) Problem solving, coordination and
  management.
• Manual (M) Truck drivers, security guards,
  waiters, janitors
• Roughly correspond to Middle, High and Low
  skill jobs.
• Four assumptions
• Computers a close substitute for human labor in
  Routine tasks.
• Routine task input a complement to Abstract tasks
  
• Computers largely orthogonal to Manual tasks
• Workers ability to engage in tasks depends on
  education
• College workers do Abstract tasks.
• High-School workers do Routine or Manual tasks

46
Effect of Falling Computer Price on Factor Inputs

• Routine task demand ? (own-factor demand downward
  slope)
• Could be supplied by either
• High School workers
• New computer capital
• Will be supplied by computer capital
• High School workers will self-select into Manual
  jobs as Routine wage falls
• Routine task input rises Supplied by machines
• More workers in Manual tasks
• Modest extension education supplies respond with
  more going to college and more workers moving
  into abstract tasks

47
Summary Why does Computerization Cause
Polarization?

• Computers do not directly substitute for
  lowest-skilled workers
• Displace a set of workers in middle skilled
  routine tasks.
• International outsourcing (offshoring) likely to
  have similar impacts
• Leads to polarization through three channels
• Directly lowers the wage of middle-skill tasks.
• Raises the wage of high-skilled (Abstract) tasks
  through Q-complementarity.
• Can raise or lower wage of low-skilled (Manual)
  tasks
• Raises wage through Q-complementarity.
• Reduces wage through labor supply impact (workers
  displaced from Routine tasks).

48
Overall Conclusions

• Evolution of U.S. Wage Structure, 1890-2005
• -- Substantial Narrowing, 1910-1950
• -- Substantial and Monotonic Widening, 1980s
• -- Widening with Polarization, 1990s 2000s
• Driving Forces Race between Demand (SBTC) and
  Supply of Skills (Education)
• Key Puzzles and Questions
• -- Discontinuous nature of narrowing (WWI and
  WWII) with persistence (institutions, shocks kept
  in place by supply growth)
• -- Why slowdown in growth of educational
  attainment? Are we back on track?
• -- Will polarization continue? Role of
  outsourcing?