Job Polarization and Structural Change Zs´ofia L. B´ar´any Sciences Po
and
Christian Siegel University of Exeter
December 2014
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
1 / 62
Job polarization is a widely documented phenomenon in developed countries since the 1980s: employment has been shifting from middle to low- and high-income occupations average wage growth has been slower for middle-income occupations than at both extremes Main explanation: routinization hypothesis In this paper 1
we document a set of facts → routinization is not the sole driving force behind this phenomenon
2
based on these facts we propose a novel perspective on the polarization of labor markets → one based on structural change
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
2 / 62
This paper empirically 1
shows that polarization has started as early as the 1950-1960s → started before ICT or increased trade could have impacted the labor market
2
shows that polarization is also present in terms of broadly defined sectors: low-skilled services, manufacturing, high-skilled services
3
shows that a significant part of occupational employment polarization is driven by employment shifts between industries
Observing that 1
polarization seems to be a long-run phenomenon
2
middle earning jobs are in manufacturing
3
the structural shift from manufacturing to services started in the 1950-1960s
→ is structural change driving the polarization of the labor market? B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
3 / 62
This paper theoretically constructs a parsimonious model where the joint analysis of wages and employment is possible Roy-type selection mechanism in a multi-sector growth model if the elasticity of substitution across goods is non-unitary a change in relative productivities triggers a change in labor demand across sectors wages in expanding sectors have to increase three sectors: manufacturing, low- and high-skilled services splitting services in two is driven by production and consumption side considerations
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
4 / 62
This paper theoretically proposes a structural change driven explanation for the joint polarization of employment and wages if the goods and the two types of services are complements then as relative productivity in manufacturing increases labor has to reallocate from manufacturing to both services in order to meet the demands to attract more workers their wages have to improve relative to manufacturing manufacturing jobs tend to be in the middle ⇒ polarization pattern
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
5 / 62
This paper quantitatively
we calibrate the model to 1950 data calculate sectoral labor productivity growth we find that our model predicts around 2/3 of the relative average wage gain of high- and low-skilled services compared to manufacturing our predictions in terms of employment share are within a 20 percent range of the data
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
6 / 62
Roadmap
1
Literature
2
Empirical evidence
3
Model
4
Quantitative Results
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
7 / 62
100 x Change in Employment Share 0 .1 .2 .
Evidence – polarization
US: Autor, Katz, Kearney (2006); Acemoglu, Autor (2010); Autor, Dorn (2013); UK: Goos, Manning (2007), Germany: Dustmann, Ludsteck, Schonberg (2009), Europe: -.1
Spitz-Oener (2006); Goos, Manning, Salomons (2009, 2014); Michaels, Natraj, Van -.2
Reenen (2013)
0
20
Panel A
40 60 80 Skill Percentile (Ranked by Occupational Mean Wage)
100
Panel B
Smoothed Changes in Employment by Skill Percentile 1980-2005
-.2
.05
-.1
.1
Change in Real Log Hourly Wage .15 .2 .25
100 x Change in Employment Share 0 .1 .2 .3
.3
.4
Smoothed Changes in Real Hourly Wages by Skill Percentile 1980-2005
0
20
40 60 80 Skill Percentile (Ranked by Occupational Mean Wage)
Panel B
Smoothed Changes in Real Hourly Wages by Skill Percentile 1980-2005
100
0
20
40
60
80
Skill Percentile (Ranked by 1980 Occupational Mean Wage)
100
Figure 1. Smoothed Changes in Employment (Panel A) and Hourly Wages (Panel B) by Skill Percentile, 1980-2005.
.3
Source: Figure 1. from Autor, Dorn (2013) B´ ar´ any and Siegel (Sciences Po, Exeter)
JobinPolarization and Structural Change / 62 to the upper tail, modest gains in the lower tail, and substantially smaller8gains
Proposed explanations for polarization
routinization: information and communication technologies (ICT) displace labor in tasks that can be described as routine Autor, Levy, Murnane (2003); Goos, Manning (2007); Michaels, Natraj, Van Reenen (2013); Autor, Dorn (2013); Goos, Manning, Salomons (2014)
off-shoring: middle earning occupations tend to be more off-shorable Grossmann, Rossi-Hansberg (2008); Blinder and Krueger (2009); Goos, Manning, Salomons (2014)
demand effects: the rise in the share of income going to the rich increases the demand for low-skilled service workers Manning (2004); Mazzolari, Ragusa (2013)
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
9 / 62
Evidence – structural change Employment shares in 3 US sectors 0.8 0.7
services
0.6 0.5
manufacturing
0.4 0.3 0.2
agriculture
0.1 0
1989
1979
1969
1959
1949
1939
1929
1919
1909
1899
1889
1879
1869
Source Figure 1. from Ngai and Pissarides (2008) B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
10 / 62
Proposed explanations for structural change focuses on sectors and on changes in output and in employment shares 1. preferences income elasticity of demand for services greater than one ⇒ changing relative demands as income increases Matsuyama (1991); Kongsamut et al (2001); Foellmi, Zweimuller (2008); Buera, Kaboski (2012)
2. technology faster TFP growth in manufacturing or different input-intensity across sectors and changing relative supply of inputs ⇒ relative sectoral prices change Ngai, Pissarides (2007); Caselli, Coleman (2001); Acemoglu, Guerrieri (2008)
→ we introduce a Roy-type selection mechanism to analyze the joint evolution of sectoral employment and wages → we distinguish between high- and low-skilled services B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
11 / 62
Roadmap
1
Literature
2
Empirical evidence
3
Model
4
Quantitative Results
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
12 / 62
Three new facts
1. polarization started as early as 1950/1960
2. it is present across broadly defined sectors
3. between industry shifts important for occupational employment
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
13 / 62
Wage polarization from Census data I.
Change in Log Wage
-.3
-.3 0
30-Yr Change in ln(real wage) 0 .3 .6 .9 1.2 1.5 1.8 2.1
30-Yr Change in ln(real wage) .3 .6 .9 1.2 1.5 1.8 2.1 2.4 2.7
Change in Log Wage
0
20 40 60 80 Occupation's Percentile in 1950 Wage Distribution 1950 - 1980 1970 - 2000
B´ ar´ any and Siegel (Sciences Po, Exeter)
100
0
1960 - 1990 1980 - 2007
Job Polarization and Structural Change
20 40 60 80 Occupation's Percentile in 1980 Wage Distribution 1950 - 1980 1970 - 2000
100
1960 - 1990 1980 - 2007
14 / 62
Employment polarization from Census data I.
Change in Employment Share
-.4
-.4
30-Yr Change in Employment Share -.3 -.2 -.1 0 .1 .2 .3 .4
30-Yr Change in Employment Share -.3 -.2 -.1 0 .1 .2 .3 .4
Change in Employment Share
0
20 40 60 80 Occupation's Percentile in 1950 Wage Distribution 1950 - 1980 1970 - 2000
B´ ar´ any and Siegel (Sciences Po, Exeter)
100
0
1960 - 1990 1980 - 2007
Job Polarization and Structural Change
20 40 60 80 Occupation's Percentile in 1980 Wage Distribution 1950 - 1980 1970 - 2000
100
1960 - 1990 1980 - 2007
15 / 62
Wage and employment polarization from Census data II. .4
Change in Log Median Wage 1980 - 2007
.8
Change in Log Median Wage 1950 - 1980 1
.7
.3
1 3
.2
8
.6
8
9
7
10
3
.1
5
.5
7
2
4
.4
0
2
9 6
10
4
2.4
2.6 Median Log Wage 1980
epanechnikov kernel
2.8
3
2.2
2.4
fractional polynomial
2.6 Median Log Wage 1980
epanechnikov kernel
9 1 7 3 8
2
3
10
0
0
8 7
3
fractional polynomial
.05
10 9
6
2.8
Change in Employment Share 1980 - 2007
.05
Change in Employment Share 1950 - 1980
2
5
-.1
6
2.2
5 1
6
-.05
-.05
5
4
-.1
-.1
4
2.2
2.4
2.6 Median Log Wage 1980
epanechnikov kernel
2.8 fractional polynomial
3
2.2
2.4
2.6 Median Log Wage 1980
epanechnikov kernel
2.8
3
fractional polynomial
classification B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
16 / 62
Wage and employment polarization from Census data III.
Employment shares of occupations
0
.5
.75
1
1.25
Share in Employment .2 .6 .4
1.5
.8
1.75
Relative average wages compared to routine jobs
1950
1960
1970
1980 Year manual
1990
2000
2010
1950
1960
abstract
1970 manual
1980 Year routine
1990
2000
2010
abstract
classification
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
17 / 62
Three new facts
1. polarization started as early as 1950/1960
2. it is present across broadly defined sectors
3. between industry shifts important for occupational employment
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
18 / 62
Wage and employment polarization across sectors
Employment shares of industries
0
.5
.75
1
Share in Employment .6 .2 .4
1.25
.8
1.5
Relative residual wages compared to manufacturing jobs
1950
1960
1970
1980 year
low-skilled services
1990
2000
high-skilled services
2010
1950
1960
1970
low-skilled serv.
1980 Year
1990
manufacturing
2000
2010
high-skilled serv.
classification regression gender and age effects
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
19 / 62
Three new facts
1. polarization started as early as 1950/1960
2. it is present across broadly defined sectors
3. between industry shifts important for occupational employment
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
20 / 62
Shift-share decomposition
∆Eot =
X
λoi ∆Eit +
X
|
∆λoit Ei
i
i
{z
B ≡∆Eot
}
|
{z
W ≡∆Eot
}
Decompose the change in an occupation’s employment share to a between industry component a within industry component
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
21 / 62
Shift-share decomposition
30-year periods
Manual
Total ∆ Between ∆ Within ∆
Employment shares 3x3 10 x 11 2.98 3.12 4.39 3.98 -1.41 -0.85
Routine
Total ∆ Between ∆ Within ∆
-19.79 -10.46 -9.33
-25.80 -12.38 -13.42
Abstract
Total ∆ Between ∆ Within ∆
16.81 6.07 10.74
19.79 8.94 10.84
Average
Total ∆ Between ∆ Within ∆
-21.16 -12.05 -9.10
-20.04 -8.83 -11.21
only within shifts
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
22 / 62
Summary of key observations
1. polarization started as early as 1950/1960 2. it is present across broadly defined sectors: low-skilled services, manufacturing, high-skilled services 3. between industry shifts important for occupational employment ⇒ structural shift of the economy might be the driving force behind the polarization of the labor market
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
23 / 62
Roadmap
1
Literature
2
Empirical evidence
3
Model
4
Quantitative Results
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
24 / 62
Key ingredients
multi-sector growth model where the output of different sectors are complements in consumption with a Roy-type selection mechanism individuals who are heterogeneous along a range of skills optimally select which sector to work in given uneven productivity growth labor has to re-allocate to slower growing sectors wages in the slower growing sectors have to increase
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
25 / 62
Production - perfect competition Low-skilled service goods Yl = Al Ll
⇒
wl = pl Al
Ym = Am Nm
⇒
wm = pm Am
Manufacturing goods
Nm – efficiency units of labor wm – wage per efficiency unit of labor High-skilled service goods Ys = As Ns
⇒
ws = ps As
note: * in producing M and S efficiency units of labor matter ⇒ income of someone with a efficiency units in M/S is awm /aws * in L raw labor is used and income is wl if working in L B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
26 / 62
Labor Supply
every individual works full time in one of the market sectors continuum of different types heterogeneity in innate ability (am , as ) ∈ R2+ for simplicity assume: I I I
am : efficiency units of labor in M → earn am wm if in M as : efficiency units of labor in S → earn as ws if in S all individuals equally productive in L → earn wl if in L
each agent chooses, given ability, the sector that provides the highest income
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
27 / 62
Sector of work decision
The earnings of an individual given wage rates wl , wm and ws with innate ability (am , as ) is wl in L, am wm in M and as ws in S. The optimal sector choice of individuals can be characterized by two cutoff values: wl wm wl abs ≡ . ws abm ≡
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
28 / 62
Sector of work decision: endogenous sorting as abs a abm m
S
M
abs
L abm
am
impact of higher abm , abs and abm /b as B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
29 / 62
Demand
The stand-in household solves: ε ε−1 ε−1 ε−1 ε−1 ε ε ε max ln θl Cl + θ m Cm + θ s Cs Cl ,Cm ,Cs
subject to s.t.
pl Cl + pm Cm + ps Cs ≤ wl Ll + wm Nm + ws Ns
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
30 / 62
Demand
The household’s optimal consumption bundle has to satisfy:
B´ ar´ any and Siegel (Sciences Po, Exeter)
Cl = Cm
Cs = Cm
pl θm pm θl
−ε
ps θm pm θs
−ε
,
.
Job Polarization and Structural Change
31 / 62
Equilibrium The equilibrium is defined as a set of cutoff sector of work abilities {b am , abs } wages per (efficiency) unit {wl , wm , ws } prices {pl , pm , ps } consumption demands {Cl , Cm , Cs } given the level of productivities {Al , Am , As } such that workers choose sector of work optimally the labor markets clear (for L, M, and S labor) the goods markets clear (for L, M, and S goods)
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
32 / 62
Structural change Using goods market clearing and market clearing wage rates in hh demand we get: am , abs ) wl Am θm −ε Al Ll (b , = A N (b a , ab ) w A θ | m{z l} l | m m{z m s } pl pm
Cl Cm
As Ns (b am , abs ) = A N (b a , ab ) | m m{z m s } Cs Cm
ws Am θm w A θ | m{z s} s
−ε
ps pm
a change in relative productivities has two direct effects: demand and supply
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
33 / 62
Structural change
Using optimal sector of work cutoffs: Ll (b am , abs ) (b am ) ε = Nm (b am , abs ) Ns (b am , abs ) Nm (b am , abs )
abm abs
ε
=
Am Al
1−ε
Am As
1−ε
θm θl
−ε
θm θs
−ε
,
.
These two equations implicitly define abm , abs , which fully characterize the equilibrium.
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
34 / 62
Structural change Proposition When manufacturing goods and the two types of services are complements (ε < 1), then faster productivity growth in manufacturing than in both types of services (dAm /Am > dAs /As = dAl /Al ), leads to a change in the optimal sorting of individuals across sectors. In particular abm unambiguously increases, abs can rise or fall, d abm /b am > d abs /b as . This leads to an increase in employment in L, an increase in effective employment in S, a fall in effective and raw employment in M.
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
35 / 62
Structural change – optimal sorting
as
abs a abm m
abs a abm m
as
abs a abm m
abs a abm m
abs abs
abs abs
abm abm
B´ ar´ any and Siegel (Sciences Po, Exeter)
am
abm
Job Polarization and Structural Change
abm
am
36 / 62
Structural change – relative average wages
Low-skilled service relative to manufacturing: wl = wm
wl wm N m Lm
=
abm wl 1 = . N m wm L am m
High-skilled service relative to manufacturing: ws = wm
B´ ar´ any and Siegel (Sciences Po, Exeter)
ws N s Ls wm N m Lm
ws = wm
Ns Ls Nm Lm
=
abm as . abs am
Job Polarization and Structural Change
37 / 62
Structural change – relative value added
Proposition When manufacturing goods and the two types of services are complements (ε < 1), then faster productivity growth in manufacturing than in both types of services (dAm /Am > dAs /As = dAl /Al ), increases the relative value added in both high- and low-skilled services compared to manufacturing: pl Yl ps Ys > 0 and d > 0. d pm Ym pm Ym
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
38 / 62
Structural change – relative value added Since pi Yi = pi Ai Ni = wi Ni , relative value added shares can be expressed as: abm Ns ps Ys ws N s = = , pm Ym w m Nm abs Nm pl Yl wl Ll Ll = = abm . pm Ym w m Nm Nm
Moreover, since wi Ni = w i Li , relative VA can be expressed as: w i Li pi Yi = . pj Yj w j Lj
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
39 / 62
Roadmap
1
Literature
2
Empirical evidence
3
Model
4
Quantitative Results
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
40 / 62
Calibration Strategy data targets: I
I
we categorize workers as low-skilled service, manufacturing, or high-skilled service based on their industry code (ind1990) in the Census/ACS four key moments of the US economy in 1950 F F
sectoral employment shares (hours worked) relative average sectoral wages
all parameters are time-invariant only exogenous change over time is productivity growth I
similarly to Ngai and Petrongolo (2014) we calculate labor productivity growth F
F I
I
by dividing sectoral value added output data from Herrendorf, Rogerson, Valentinyi (2013) with sectoral employment data from the BEA
due to data limitations we cannot break the labor productivity growth of services into low- and high-skilled possibilities: raw/adjusted, average/decennial
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
41 / 62
Calibration of ability distribution
assume f (am , as ) is uniform normalize the mean of am and as to be unity (not separately identified) → need to find am and as given f (am , as ), the observed labor shares uniquely identify the sector-of-work cutoffs, the sector-of-work cutoffs in turn imply relative average wages → pin down am , as such that when matching the raw employment shares in 1950, the model also matches the relative average wages
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
42 / 62
Calibration of utility function the parameters of the utility function ε, θl , θm , θs take ε, the elasticity of substitution from the literature I
ε has been estimated by Herrendorf, Rogerson, Valentinyi (2013); when sectoral output is measured in value added terms, ε = 0.002
I
Ngai and Pissarides (2008) find that plausible estimates are in the range [0, 3]
calibrate τl =
θm θl
−ε
and τs =
θm θs
−ε
to match 1950 relative average wages and employment shares
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
43 / 62
Calibrated parameters
[am , am ] [as , as ] ε τl τs
Description range of manufacturing efficiency range of high-skilled service efficiency CES b/w L, M and S in consumption relative weight on M relative weight on S
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
Value [0.6292, 1.3708] [0, 2] 0.002 0.1752 0.5424
44 / 62
Adjustment for average labor efficiency changes due to the self-selection of individuals into sectors expanding sectors increase by soaking up relatively less efficient workers contracting sectors decrease by shedding relatively less efficient workers ⇒ average efficiency of labor in expanding sectors fall, while in contracting sectors it increases ⇒ manufacturing productivity growth might be overestimated services productivity growth might be underestimated when calculating from raw employment data pointed out in the context of measuring productivity growth across sectors by Young (2014 AER), estimated for the bias in skill premium estimates by Carneiro and Lee (2011 AER) B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
45 / 62
Adjustment based on our calibration
use calibration for efficiency distribution, f (am , as ) take raw employment shares from the data given cutoff structure in our model calculate the change in average labor efficiency in each sector overall efficiency gain in manufacturing: 4.8% overall efficiency loss in services: 3.4% adjust the annual change in raw employment by calculated annual labor efficiency gain/loss in the sector → adjusted labor productivity growth
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
46 / 62
Annual average labor productivity growth
1950-1960 1960-1970 1970-1980 1980-1990 1990-2000 2000-2007 1950-2007
Based on raw Manufacturing 1.0199 1.0213 1.0115 1.0260 1.0345 1.0229 1.0226
B´ ar´ any and Siegel (Sciences Po, Exeter)
labor Services 1.0105 1.0103 1.0086 1.0002 1.0068 1.0122 1.0079
Adjusted by average efficiency Manufacturing Services 1.0213 1.0114 1.0204 1.0115 1.0103 1.0089 1.0242 1.0003 1.0333 1.0073 1.0215 1.0128 1.0218 1.0085
Job Polarization and Structural Change
47 / 62
Transition under baseline productivity growth Index of sectoral productivities Am 3.5
ba m ba s
1.15
A, A l
Sector of work cutoffs
1.2
4
s
1.1 1.05
3
1 2.5 0.95 0.9
2
0.85 1.5 0.8 1 1950
1960
1970
1980
1990
2000
2010
0.75 1950
Employment shares: data vs model
1960
1970
1980
1990
2000
2010
Relative average wages: data vs model
0.7
1.4 Ll
wl / wm
Lm
0.6
Ls
1.3
ws /wm
1.2
0.5
1.1 0.4 1 0.3 0.9 0.2
0.8
0.1 0 1950
0.7
1960
1970
B´ ar´ any and Siegel (Sciences Po, Exeter)
1980
1990
2000
2010
1950
1960
1970
Job Polarization and Structural Change
1980
1990
2000
2010
48 / 62
Transition under adjusted productivity growth
Relative average wages: data vs model
Employment shares: data vs model 0.7
1.4 Ll
wl /wm
Lm
0.6
1.3
ws / wm
Ls 1.2 0.5 1.1 0.4 1 0.3 0.9 0.2 0.8 0.1
0 1950
0.7
1960
1970
1980
B´ ar´ any and Siegel (Sciences Po, Exeter)
1990
2000
2010
1950
1960
Job Polarization and Structural Change
1970
1980
1990
2000
2010
49 / 62
Transition under decennial productivity growth Index of sectoral productivities
Sector of work cutoffs 1.2
4 Am 3.5
ba m ba s
1.15
Al, As
1.1 1.05
3
1 2.5 0.95 0.9
2
0.85 1.5 0.8 1 1950
1960
1970
1980
1990
2000
2010
0.75 1950
Employment shares: data vs model
1960
1970
1980
1990
2000
2010
Relative average wages: data vs model
0.7
1.4 Ll
wl / wm
Lm
0.6
1.3
ws /wm
L
s
1.2
0.5
1.1 0.4 1 0.3 0.9 0.2
0.8
0.1 0 1950
0.7
1960
1970
B´ ar´ any and Siegel (Sciences Po, Exeter)
1980
1990
2000
2010
1950
1960
1970
Job Polarization and Structural Change
1980
1990
2000
2010
50 / 62
Value added shares Relative manufacturing value added: data vs model 2.4 2.2
2
1.8 1.6
1.4
1.2 1
0.8
0.6
0.4 1950
1960
B´ ar´ any and Siegel (Sciences Po, Exeter)
1970
1980
1990
Job Polarization and Structural Change
2000
2010
51 / 62
Robustness Relative average wages: data vs model
Employment shares: data vs model 0.7
1.4 Ll
wl / wm
Lm
0.6
1.3
ws / wm
Ls 1.2 0.5 1.1 0.4 1
* ε = 0.02
0.3 0.9 0.2 0.8 0.1
0 1950
0.7
1960
1970
1980
1990
2000
2010
1950
1960
1970
1980
1990
2000
2010
Relative average wages: data vs model
Employment shares: data vs model 0.7
1.4 Ll
wl / wm
Lm
0.6
1.3
ws / wm
Ls 1.2 0.5
* ε = 0.2
1.1 0.4 1 0.3 0.9 0.2 0.8 0.1
0 1950
B´ ar´ any and Siegel (Sciences Po, Exeter)
0.7
1960
1970
1980
1990
2000
2010
Job Polarization and Structural Change
1950
1960
1970
1980
1990
2000
2010
52 / 62
Summary from data 1. polarization started as early as the 1950s 2. polarization present also across sectors 3. between industry shifts important for occupational employment → structural change possible force driving polarization the model: we introduce heterogeneous labor via Roy-type selection into a multi sector model of growth unbalanced technological change not only leads to the re-allocation of employment and expenditure shares but also affects sectoral average wages quantitatively: this simple model does remarkably well in matching the labor market patterns of the last 60 years B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
53 / 62
Thank you!
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
54 / 62
Occupation categories The 10 occupational codes are: 1. personal care; 2. food and cleaning services; 3. protective services; 4. operators, fabricators and laborers; 5. production, construction trades, extractive and precision production; 6. administrative and support occupations; 7. sales; 8. technicians and related support occupations; 9. professional specialty occupations; 10. managers.
back
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
55 / 62
1
Manual: low-skilled non-routine housekeeping, cleaning, protective service, food prep and service, building, grounds cleaning, maintenance, personal appearance, recreation and hospitality, child care workers, personal care, service, healthcare support;
2
Routine construction trades, extractive, machine operators, assemblers, inspectors, mechanics and repairers, precision production, transportation and material moving occupations, sales, administrative support, sales, administrative support; sales, administrative support
3
Abstract: skilled non-routine managers, management related, professional specialty, technicians and related support.
back
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
56 / 62
Industry classification 1
Low-skilled services personal services, entertainment, business and repair services (except advertising and computer and data processing services), nursing and personal care facilities, child day care service, family child care homes, residential care facilities, social services not elsewhere classified, taxi, retail bakeries, eating and drinking places;
2
Manufacturing mining, construction, manufacturing, transport and public utilities, wholesale trade, retail trade;
3
High-skilled services communications, finance, insurance and real estate, theaters and motion pictures, professional and related services, public administration, advertising, computer and data processing services.
back
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
57 / 62
Regression of log hourly wages: industry effects
Year LS HS Cont Obs R2
1950 -0.36 (0.01) 0.03 (0.00) yes 113635 0.18
1960 -0.37 (0.00) 0.09 (0.00) yes 459564 0.19
1970 -0.26 (0.00) 0.16 (0.00) yes 579290 0.16
1980 -0.26 (0.00) 0.11 (0.00) yes 958318 0.19
1990 -0.24 (0.00) 0.19 (0.00) yes 1094458 0.21
2000 -0.20 (0.00) 0.22 (0.00) yes 1235282 0.19
2007 -0.22 (0.00) 0.28 (0.00) yes 1308885 0.22
Standard errors in parentheses
controls: a polynomial in potential experience (defined as age - years of schooling - 6), dummies for gender, race, and born abroad back
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
58 / 62
Descriptive statistics
HSD HSG SColl Coll PG Educ Yrs Female Foreign-Born
low-skilled services 23.79% 34.95% 27.90% 10.35% 3.01% 12.21 54.12% 16.18%
manufacturing 23.64% 38.02% 24.47% 10.94% 2.93% 12.19 27.59% 10.28%
high-skilled services 7.41% 22.81% 28.86% 24.07% 16.85% 14.21 54.72% 8.97%
back
B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
59 / 62
Gender and age effects in industry employment shares fix industry shares of gender-age cells at their 1950 level, allow age and gender shares to follow their actual path Counterfactual: Only Gender-Age Cells Vary
0
Share in Employment .2 .6 .4
.8
vary size of gender-age cells, fixing their industry shares of 1950
1950
1960
1970 L actual L cntrf
1980 Year
1990
M actual M cntrf
2000
2010
S actual S cntrf
back B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
60 / 62
Shift-share decomposition for 30-year periods 3 occ x 3 sec 50-80 80-07 50-07
10 occ x 11 ind 50-80 80-07 50-07
Manual
∆ B∆ W∆
-2.11 0.50 -2.61
5.09 3.49 1.59
2.98 4.39 -1.41
-2.08 -0.40 -1.68
5.21 4.09 1.12
3.12 3.98 -0.85
Routine
∆ B∆ W∆
-7.59 -3.41 -4.18
-12.20 -6.51 -5.69
-19.79 -10.46 -9.33
-10.94 -4.23 -6.70
-14.68 -6.84 -7.84
-25.80 -12.38 -13.42
Abstract
∆ B∆ W∆
9.70 2.90 6.79
7.11 3.02 4.09
16.81 6.07 10.74
10.65 4.17 6.47
8.93 3.76 5.17
19.79 8.94 10.84
Average
T∆ B∆ W∆
-9.37 -4.61 -4.76
-11.79 -6.67 -5.12
-21.16 -12.05 -9.10
-11.05 -4.07 -6.98
-9.00 -3.82 -5.18
-20.04 -8.83 -11.21
back B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
61 / 62
How important are within-industry shifts in occ shares? fix industry shares at 1950 level, within-ind occ shares follow the actual path → compare to actual overall occupational share paths Counterfactual: Only Within Industry Shift of Occupations
0
Share in Employment .2 .4 .6
.8
fix shares of industries, vary occupation shares within
1950
1960
1970 manual cntrf
1980 Year
1990
routine cntrf
2000
2010
abstract cntrf
back B´ ar´ any and Siegel (Sciences Po, Exeter)
Job Polarization and Structural Change
62 / 62
