Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
A Tale of Two Countries:A Story of the French and US Polarization
Julien AlbertiniUniversite de Lyon 2 (GATE)
Jean-Olivier HairaultUniversite de Paris I & PSE & Cepremap
Francois LangotLe Mans Universite (GAINS-TEPP, IRA) & Cepremap & PSE
Thepthida SopraseuthUniversite de Cergy-Pontoise (Thema) & Cepremap
June 2018, DARES, Paris
The objective of the paper
I Identifying the main driving forces underlying jobpolarization in France and in the US
I Job polarization is usually defined as a growing proportion ofjobs concentrated at the tails of the skill or wage distribution
I Change in the shares of three types of tasks
1. decreasing share of routine tasks (blue-collar and clerkworkers...),
2. increasing share of abstract, cognitive jobs (engineers, lawyers,professionals ....)
3. increasing share of manual services (health care, household andhousework services, ...)
Due to Task-Biased Technological Change (TBTChypothesis), and not Skill-Biased TC.
1
The objective of the paper
Apparently the same polarization in France and in the US, alreadydocumented by Goos, Manning, Salomons (2009).
II But actually not driven by the same forces
1. Labor Market Institutions (LMI) in France2. TBTC in the US
Leading to very different aggregate outcome in terms ofemployment rate and wage inequality
II The same polarization in the data is actually very different interms of employment and welfare performance.
2
Job polarization captured by trends in employment shares
1985 1990 1995 2000 2005
0.28
0.30
0.32
0.34
0.36
a. Abstract
0.15
0.18
0.21
0.24
US (left scale)France (right scale)
1985 1990 1995 2000 20050.45
0.50
0.55
b. Routine
0.65
0.70
0.75
1985 1990 1995 2000 2005
0.15
0.16
0.17
c. Manual
0.09
0.10
0.11
Authors’ calculations. Based on BLS and French LFS.
US data French data
2
Employment per capita in the US and France
1985 1990 1995 2000 2005
0.16
0.18
0.20
0.22
a. Abstract
0.07
0.08
0.09
0.10
0.11US (left scale)France (right scale)
1985 1990 1995 2000 2005
0.29
0.30
0.31
0.32
0.33
0.34
b. Routine
0.30
0.32
0.34
0.36
1985 1990 1995 2000 2005
0.09
0.10
0.11
c. Manual
0.04
0.045
0.05
1985 1990 1995 2000 2005
0.56
0.58
0.60
0.62
0.64
0.66
d. Aggregate employment
0.42
0.44
0.46
0.48
0.50
OECD Employment outlook. Employment/Population
3
Structural changes over the period : Technology
1975 1980 1985 1990 1995 2000 20050.7
0.75
0.8
0.85
0.9
0.95
1
1.05
USFrance
4
Source: Investment Price (relative toconsumption price). World Development
Indicators. The relative price is normalizedto one in 1980.Source : Data from
Karabarbounis Neiman (QJE 2014).
Structural changes over the period : Education
1960 1970 1980 1990 2000 2010 20200
0.05
0.1
0.15
0.2
0.25
0.3
0.35
USFrance
5
Source: French census. Diplomeuniversitaire 1er, 2eme ou 3eme cycle,
BTS-DUT. US Census (Years of SchoolCompleted) College 4 years and more
Structural changes over the period : Labor MarketInstitutions (LMI)
1960 1970 1980 1990 20000
0.1
0.2
0.3
0.4
0.5Replacement Ratio
RR is growing over time
USFrance
1960 1970 1980 1990 20000.1
0.2
0.3
0.4
0.5
0.6Workers bargaining power
Divergent evolution of BP
1960 1970 1980 1990 20000
0.05
0.1
0.15
0.2
0.25
0.3
0.35Employer payroll tax
Fall in payroll tax
1960 1970 1980 1990 20000.02
0.04
0.06
0.08
0.1
0.12
0.14Worker Social Security Contribution
1960 1970 1980 1990 20000.5
1
1.5
2
2.5
3
3.5Minimum wage (real hourly wage)
decline in MW
6
Structural changes over the period : Labor MarketInstitutions (LMI)
I ”usual suspects” : Minimum wage ; UB’s replacement rate ;workers’ bargaining power
I Industry-specific wage floor : a wage floor is set for everyposition and workers cannot be paid below theindustry-specific wage floor associated with their job position(Fougere, Gautier and Roux (2008)).
I At the firm level, employers and unions bargain on wageincreases provided that wages are set above the industrywage floors.
I On average, a 1% increase of the real MW raises wage floorsby about 0.3 pp and wage floor adjustments are much moreresponsive to MW variations when wage floors are close to theMW.
7
The objective of the model
I To quantify the contribution of each trend in the polarizationprocess.
I To compare France and the US
I To assess the impact of each trend on wage inequality
8
A structural approach
I Model : Autor and Dorn meet DMPMulti-sectorial search and matching model with occupationalchoice and 3 exogenous trendsI Task-Biased Tech. Change (TBTC), Autor and Dorn, 2013
I Fall in price of IT capital
I Growth in supply of skilled laborI the share of skilled workers rise
I Evolution of Labor Market Institutions :I LMIs more or less generous
I Structural estimation : using a just-identified system,I elasticity of substitution between goods in preferences, AND
between inputs in production,I trends in IT prices, and educational attainments.
9
Contribution to the literatureI SaM with technological change
I Ljungqvist and Sargent (1998, 2008), Mortensen andPissarides (1998), Hornstein et al. (2007)
I Analysis that compare two steady state equilibria withoutworker mobility choices between occupations.
I SaM with occupational choicesI Alvarez and Shimer (2011) ; Carrillo-Tudela and Visschers
(2014)I Analysis that focuses on mobility choices over the business
cycle
I Our paper :I Transitional dynamics : structural change (technological,
education and institutions gradually change) with irreversibility.I General equilibrium analysis : the demand shifts toward
services come from higher incomes of the skilled workers⇔ The diffusion process of the structural change depends onsubstitutability between consumption goods and betweenproduction factors.
10
Building blocks of the model
Service sector:
(assisting others)
Production with
unskilled labor only ��
Good sector:
Production with 2
complements :
skilled labor �� and
routine tasks
(CES basket of unskilled
labor �� and capital �,
Highly substituable)
Capital
whose price
falls due to
technological
progress
Consumers’ Demand for goods and services :
Consume CES basket of service and goods(Skilled and unskilled,
Employed and non employed consumers)
Search and matching Search and matching
↓ �
High Skill workers Low Skill worker
Occupational choice
Technological change
General equilibrium : Endogenous relative price of service
10
Occupational mobility
Employed unskilledworkers
routine tasks
Employed unskilledworkers
manual tasks
Employed unskilledworkers
learning process
Unemp. unskilledworker searchingfor routine jobs
Unemp. unskilledworker searching for
manual jobs
Unemp. unskilledworker searchingfor manual jobs
Occupationalchoice
A. SKILLED WORKERS
Employed skilledworkers
Abstract tasks
Unemployed skilledworkers
Abstract tasks
B. UNSKILLED WORKERS
« Novice » , « inexperienced »« bridge jobs »« New/old movers »
« movers », « switchers »
« experienced »
Unemployed
« stayers »
Mobility
11
Firms : Goods sector
The problem of a representative firm in the goods sector is :
Πg = max
{Y g − pkK −
∑ηηS
wr (η)ηLr (η)− waLa
−cVa − c∑η
ηSVr (η) + βΠg
+1
}
s.t. Y g ≥ ALαa
(1− µ)
η∑ηS
ηLr (η)
σ
+ (µK )σ
1−ασ
Lr ,+1(η) = (1− s)Lr (η) + qr (η)Vr (η)
La,+1 = (1− s)La + qaVa
Lr (η) ≥ 0
The last equation determines the endogenous separations.
12
Firms : Service sectorThe problem of a representative firm in the service sector is :
Πs = max
{psY
s − wmLm −∑
η wnm(η)Lnm(η)− wo
mLom
−cVm − c∑
η Vnm(η)− cV o
m + βΠs+1
}
s.t. Y s ≥ As
(Ls + δ
∑η
Lnm(η) + δLom
)Lm,+1 = (1− s)Lm + qmVm + (1− s)λ
∑η
Lnm(η) + (1− s)λLom
Lom,+1 = (1− s)(1− λ)Lom + qomVom
Lnm,+1(η) = (1− s)(1− λ)Lnm(η) + qnm(η)V nm(η)
Ls ≥ 0
δ ∈ (0, 1) the loss of efficiency due to the learning process.The last equation determines the endogenous separations.
13
Occupational choice : Ut+1 = max{Ur ,t+1,Unm,t+1}
Employed : for each ability level η
Wr ,t = wr ,t(1− τwt ) + β
[IFUt+1 + (1− IF )[(1− s)Wr ,t+1 + sUt+1]
]W n
m,t = wnm,t(1− τwt ) + λ[(1− s)βWm,t+1 + sβUm,t+1]
+(1− λ)[(1− s)βW nm,t+1 + sβUo
m,t+1]
Unemployed : for each ability level η
Ur ,t = zr + β
[(1− fr ,t)Ut+1 + fr ,tWr ,t+1
]Unm,t = zr + β
[(1− f nm,t)U
nm,t+1 + f nm,tW
nm,t+1
]fi : job finding rate, s : separation rate, λ : learning parameter, andzr the unemployment benefits.Value functions : Employees Unemployed workers
14
Occupational choices and LMIs
If LMI push up the reservation wage, these bridge jobs,poorly productive, can not be open by firms. Why ?
I UB (zr ) are determined by workers’ past earnings (LS, 1998,2008)
I When they switch, workers are eligible to an UB indexed ontheir previous routine job wage (new mover/switcher), whichcan be higher than their wage on a bridge job.
I After a long unemployment spell, or if they are fire from theirbridge jobs, they lose their eligibility on the routine UB.
I LMIs can lead to non-existence of these bridge jobsI through a strong indexation of the UB on past wages,I or through a high MW,
hence stalling labor reallocation
15
Wage setting
I Nash bargaining
I the wage wNash is highly flexible and follows both productivityand labor market tightness.
I wage bargaining takes into account occupational switch andchanges in taxation over time More
I Wage floor : real wage rigidities : the WS curve is (for eachability level η)
wt = max{(1 + x)ϑmw t ,wNasht } if mw0 < wNash
0
with w0 and wm0 is the initial wage and MW, x =wNash
0mw0
− 1and ϑ accounts for the indexation of the wage floor on theMW.
16
General equilibrium settingI Demand for goods and services from household i = a, r ,m
in the economy. They are complements in households’preferences
Ci =[νC ρg ,i + (1− ν)C ρs,i
] 1ρ
⇒ endogenous relative price of services (ps) : ”GeneralEquilibrium effect”⇒ rising ps drives surplus of manual jobs upward⇒ more manual jobs are open⇒ a signal to switch occupation to manual jobs
I General equilibrium :
Y g = Cg + pkK + cVa + c
η∑ηS
Vr (η)
Y s = Cs + cVm + c∑η
V nm(η) + cV o
m
17
Quantitative analysis :Accounting for job polarization
in the US and in France
17
Identification of the structural parameters (1)I The calibration is quarterly.I Common to both countries Φ1, with dim(Φ1) = 12
Φ1 = {β, ρ, ν, σ, µ, α, η, η,A,As , δ, ψ}I For the labor market, the parameters Φ2 are country-specific,
with dim(Φ2) = 24
Φ2 = {Υa,Υr ,Υm, sa, sr , sm, ca, c , ξa, ξr , ξm, λ}US,FI Country-specific technological change and the drift in the
supply of skilled labor :
Φ3 = {pk(0), ϑpk , pk(T ), La(0), ϑLa, La(T )}US ,F dim(Φ3) = 12,
where x = pk , La evolve as follows :
x(t) =
{x(0) if t < tx0
x(T ) + (x(0)− x(T )) exp(−ϑx(t − tx0)2) Otherwise
I The total number of parameters is dim(Φ) = 48.
18
Identification of the structural parameters (2)
I Empirical target : ΨT , for i = US ,F , with dim(ΨT ) = 28
ΨT =
Na,i (0),Nr ,i (0),Nm,i (0),Na,i (T ),Nr ,i (T ),Nm,i (T )Ei [Na],Ei [Nr ],Ei [Nm],D5,i (0)/D1,i (0),D9,i (T )/D5,i (T )Ei [D5,i/D1,i ],Ei [D9,i/D5,i ]
I In order to identify the unknown parameters, it is necessary to
introduce48︸︷︷︸
dim(Φ)
− 28︸︷︷︸dim(Ψ)
= 20 restrictions
19
Identification of the structural parameters (2)
I External information : we calibrate Φci ∈ Φi (18 restrictions)
Φc1 = {β, µ, ν, ψ} with dim(Φc
1) = 4
Φc2 = {sa, sr , sm, ca, c}US,F with dim(Φc
2) = 10
Φc3 = {La(0), La(T )}US,F with dim(Φc
3) = 4
I Assumptions (2 restrictions)
pk,US(0) = pk,F (0) pk,US(T ) = pk,F (T )
⇒ 20 restrictions : Just-identified systemI Over-identifying test :
I Does the model generate the hump shaped patterns of theemployment dynamics by occupation ?
I Does it match the dynamics of wage inequalities ?
20
Model versus data : Fitting non-linear changes in percapita employment
21
Model versus data : US wage inequality
1985 1990 1995 2000 20050.15
0.2
0.25
0.3
0.35
0.4Gini: US
DataModel
1985 1990 1995 2000 20051.2
1.4
1.6
1.8
2
2.2
2.4
2.6
2.8Wage D9/D5
1985 1990 1995 2000 20051.85
1.9
1.95
2
2.05
2.1
2.15Wage D5/D1
1985 1990 1995 2000 20053
3.5
4
4.5
5Wage D9/D1
Data : CPS-MORG. Authors’ calculations.
22
Model versus data : FR wage inequality
1985 1990 1995 2000 20050.15
0.2
0.25
0.3
0.35
Gini: FR
DataModel
1985 1990 1995 2000 20051.5
2
2.5
3Wage D9/D5
1985 1990 1995 2000 20051
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
Wage D5/D1
1985 1990 1995 2000 20051.5
2
2.5
3
3.5
4Wage D9/D1
Data : Breda et al. (2016). Gini from Piketty (2001)
23
Date of moving and scrapping
23
η = 35
23
η = 36
23
η = 37
23
η = 38
23
η = 43
23
Counterfactuals
I Objective : Assessing the role of each exogenous trend(TBTC, supply of skilled workers, LMI) in accounting for thejob polarization process.
I Benchmark model : with TBTC, LMI and educationalattainment.
I Counterfactual experiments : Cancel the evolution of 1exogenous trend (set at 1975 and constant)I 2 other trends are at work
24
USTBTC is driving job polarization, fostered by flexible LMI.
1980 1990 20000.38
0.39
0.4
0.41
0.42
0.43
b. Unskilled employment
1980 1990 20000.54
0.56
0.58
0.6
0.62
0.64
c. Aggregrate employment
1980 1990 2000
0.16
0.18
0.2
0.22
0.24
d. Abstract employment
1980 1990 2000
0.29
0.3
0.31
0.32
0.33
0.34
e. Routine employment
1980 1990 20000.085
0.09
0.095
0.1
0.105
0.11
0.115
f. Manual employment
Benchmark
Constant LMI
Constant TBTC
Constant La
25
FranceMajor impact of LMI on unskilled workers, such that TBTC cannotaffect employment. Larger role for education.
1980 1990 20000.25
0.3
0.35
0.4
0.45
0.5
b. Unskilled employment
1980 1990 2000
0.35
0.4
0.45
0.5
0.55
0.6
c. Aggregrate employment
1980 1990 20000.06
0.07
0.08
0.09
0.1
0.11
0.12
d. Abstract employment
1980 1990 20000.25
0.3
0.35
0.4
0.45
e. Routine employment
1980 1990 20000.038
0.04
0.042
0.044
0.046
0.048
0.05
0.052
f. Manual employment
Constant TBTC
Benchmark
Constant LMI
Constant La
26
France : Payroll tax
1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 20150.14
0.16
0.18
0.2
0.22
0.24
0.26
0.28
0.3
0.32
0.34
Total (data)High-skilledLow-skilled
27
France : The rebound of routine jobs in the late 1990s dueto tax exemption
28
Counterfactual : Wage inequality in the US
I TBTC increases the top-wages (large gains for abstract), butreduces the inequalities at the bottom (catch-up of themanual jobs)
I LMI changes magnify the rise in wage inequality during the80s.
29
Counterfactual : Wage inequality in France
I TBTC increases the top-wages (abstract jobs), but does notaffect the bottom of the wage distribution
I LMI changes compress the wage distribution at the bottom.
30
Conclusion
The impact of task biased technological change, labor marketinstitutions, and rising educational attainment on job polarizationI US : gains in aggregate employment.
I TBTC accounts for 50% of rise in aggregate employmentI fostered by de-unionization
I France : Losses in aggregate employmentI Without changes in LMI, aggregate employment would have
risenI In this context, TBTC cannot boost employment
I Inequality :I US : TBTC accounts for 2/3 of the rise in wage inequalityI French LMI dampens the rise in wage inequality (the reverse in
the US)
31
Appendix
31
US : CPS monthly, 1982m01-2017m08
I Abstract → Abstract : 95.6%
I J2J :
{Abstract → Routine = 1.4%Abstract → Manual = 0.3%
}= 38% of turnover
I Abstract → Unemployment + Out of LF : 2.7%
I Routine → Routine : 93.6%
I J2J :
{Routine → Abstract = 1%Routine → Manual = 0.6%
}= 25% of turnover
I Routine → Unemployment + Out of LF : 4.8%
I Manual → Manual : 89.8%
I J2J :
{Manual → Abstract = 0.9%Manual → Routine = 2.0%
}=28% of turnover.
I Manual → Unemployment + Out of LF : 7.3%
31
French LFS quarterly, 2003Q1-2016Q4
I Abstract → Abstract : 97.2%
I J2J :
{Abstract → Routine = 0.2%Abstract → Manual = 0%
}= 7% of turnover
I Abstract → Unemployment + Out of LF : 2.6%
I Routine → Routine : 95%
I J2J :
{Routine → Abstract = 0.2%Routine → Manual = 0.1%
}= 6% of turnover
I Routine → Unemployment + Out of LF : 4.8%
I Manual → Manual : 93.7%
I J2J :
{Manual → Abstract = 0.2%Manual → Routine = 0.6%
}=12% of turnover.
I Manual → Unemployment + Out of LF : 5.5%
Back to slides
31
Figure – Omitting transitions in the US
1980 1990 2000 2010 2020
0.16
0.18
0.2
0.22
0.24
0.26EA
1980 1990 2000 2010 20200.24
0.26
0.28
0.3
0.32
0.34
0.36ER
1980 1990 2000 2010 2020
0.05
0.1
0.15
EM
steady state counterfactualsteady state
Back to slides
31
Source: Steady state employment stocksusing transition matrices from monthly
CPS.Counterfactual steady state stocks whenthe following transitions are set to zero :
AR, AM, RA,RM,MR,MA.Source : Charlot, Fontaine, Sopraseuth
(2018)
Polarization in the US : Autor & Dorn (2013)Back to slide intro
AbstractTasks= 33%
RoutineTasks= 45%
ManualTasks= 22%
31
Related literature
Our contribution : labor reallocation with occupational changes ina non-stationary environment, within unskilled workers (from themiddle towards the bottom of the wage distribution), outsidesteady state
I Job polarization as an outcome of the structural change :Autor and Dorn (2013)
I Search and matching, technological changes : Mortensen andPissarides (1998,1999), Horstein and al. (2004)
I Occupational choice - search vs rest unemployment : Alvarezand Shimer (2011)
I ”European employment problem” and the interaction betweenstructural change and LMI : Ljungqvist and Sargent (1998,2008), Blanchard and Wolfers (1999)
31
Links with the ”TC-LMI interaction” literature
Ljungqvist and Sargent (1998 & 2008), Mortensen and Pissarides(1999), Hornstein, Krussel & Violante (2007). Originality of ourpaper w.r.t this literature :
I Perfect mobility versus mobility costs → LMI
I Steady-state versus transitional dynamics (the path of LMImatters)
I In our paper,I Mobility towards less productive jobs (job polarization)I A more comprehensive view on labor market dynamics :
aggregate employment, employment by task, wage dynamicsand inequalities
I Understanding employment growth by quantifying the relativecontribution of LMI, TC and Labor supply of skilled labor
I Reform packages : stress on interaction between LMI
31
US Data, as in Jaimovich and Siu (2015) Back to slides
I Employment Data by Occupation from BLS
I Abstract : Non-routine cognitive workers. Management,business, and financial operations occupations. Professionaland related occupations.
I Routine : sales and related occupations. office andadministrative support occupations. production occupations,transportation and material moving occupations, constructionand extraction occupations, and installation, maintenance,and repair occupations.
I Manual : service occupations : ... Ushers, Lobby Attendants, and Ticket Takers ;
Amusement and Recreation Attendants ; Embalmers ; Funeral Attendants ; Morticians, Undertakers, and
Funeral Directors ; Barbers ; Hairdressers, Hairstylists, and Cosmetologists ; Makeup Artists, Theatrical and
Performance ; Manicurists and Pedicurists ; Shampooers ; Skincare Specialists ; Baggage Porters and
Bellhops ; Concierges ; Travel Guides ; Childcare Workers ; Personal Care Aides ; Fitness Trainers and
Aerobics Instructors ; Recreation Workers ; Residential Advisors ; Personal Care and Service Workers, All
Other
31
Back to slides
US Data, as in Jaimovich and Siu (2015)
I Consistent with Autor and Dorn’s classification
I Consistent with Routine-Task Intensity index based on DOT
31
Back to slides
French data :
I Annual French Labor surveys (1983-2014)
I Compute employment by occupation
I Abstract, Routine and Manual workers are identified in thesame way as in Jaimovich and Siu (2015)
I classification using wages is not possible in the early 1980s(wage is not a continuous variable in the early 1980s)
31
Back to slides Goods sector : Complementarity and substitutabilityModel : details Goods
Y g ≥ ALαa
(1− µ)
η∑ηS
ηLr (η)
σ
+ (µK )σ
1−ασ
Service sector Model : details Services
Y s ≥ As
(Ls + δ
∑η
Lnm(η) + δLom
)
Preferences : complementarities Model : details Households
C =[νC ρg + (1− ν)C ρs
] 1ρ
32
Firms : Goods sector
The representative firm’s problem
Πg = max
{Y g − pkK −
∑ηηS
wr (η)ηLr (η)− waLa
−cVa − c∑η
ηSVr (η) + βΠg
+1
}
s.t. Y g ≥ ALαa
(1− µ)
η∑ηS
ηLr (η)
σ
+ (µK )σ
1−ασ
Lr ,+1(η) = (1− s)Lr (η) + qr (η)Vr (η)
La,+1 = (1− s)La + qaVa
Πg = max{
Πg(Lr (η)>0),−FC × Lr (η) + Πg
(Lr (η)=0)
}Back to slide model
32
Firms : Service sector
The representative firm’s problem
Πs = max
{psY
s − wmLm −∑
η wnm(η)Lnm(η)− wo
mLom
−cVm − c∑
η Vnm(η)− cV o
m + βΠs+1
}
s.t. Y s ≥ As
(Ls + δ
∑η
Lnm(η) + δLom
)Lm,+1 = (1− s)Lm + qmVm + (1− s)λ
∑η
Lnm(η) + (1− s)λLom
Lom,+1 = (1− s)(1− λ)Lom + qomVom
Lnm,+1(η) = (1− s)(1− λ)Lnm(η) + qnm(η)V nm(η)
with δ ∈ (0, 1) the loss of efficiency due to the ”movers”’ learningprocess.
Back to slide model
32
Households : Demand
For each worker, the budgetary constraint is
PC = I with I ∈ {wa,wr (η),ws ,wm, za, zs , zr}
Given that all workers, we have
C =[νCρg + (1− ν)Cρs
] 1ρ P =
[ν
11−ρ + (1− ν)
11−ρ p
ρρ−1s
] ρ−1ρ
the optimal sharing of the basket good C is given by :
ps =1− νν
(Cg
Cs
)1−ρ
⇒
{Cg = ν
11−ρ(
1P
) 1ρ−1 I
P
Cs = (1− ν)1
1−ρ(psP
) 1ρ−1 I
P
which are the demand functions.Back to slide model
32
Model Assumptions : labor reallocation across sectors
Back to slides
I A mobility cost = a market for ”movers/switcher” (s) :I Some l-skill workers, unemployed on a ”routine” labor market,
can choose to move to search for a ”manual” job.I For them, the cost is the acceptance of a bad job in the
”manual” sector
I Learning process : the duration of the transformation of a badjob into a good job in the manual sector is stochastic with aPoisson parameter λ.
I There is potentially 2 types of switchers :I The first are eligible to an UB indexed on their previous
”routine” job wage : new mover/switcher.I The second have a longer experience on this segment of the
labor market and have lost their eligibility on this UB.
32
Employees’ Opportunities
The worker’s value functions are
Wa = wa(1− τwa) + (1− s)βWa,+1 + sβUa,+1
Wm = wm(1− τw ) + (1− s)βWm,+1 + sβUm,+1
Wr (η) = ηwr (η) + (1− s)βWr ,+1(η) + sβmax{Ur ,+1(η),Unm,+1(η)}
W om = wo
m(1− τw ) + λ[(1− s)βWm,+1 + sβUm,+1]
+(1− λ)[(1− s)βW om,+1 + sβUo
m,+1]
W nm(η) = wn
m(η)(1− τw ) + λ[(1− s)βWm,+1 + sβUm,+1]
+(1− λ)[(1− s)βW nm,+1(η) + sβUo
m,+1]
I ”Movers” can obtain a good ”manual” job with a Proba = λ
I For workers previously occupied on a ”Routine” task, thereallocation is an option ⇔ max{Ur ,+1(η),Un
m,+1(η)}.Back to slide model
32
Unemployed workers Opportunities
For the unemployed worker,
Ua = za + (1− fa)βUa,+1 + faβWa,+1
Um = zm + (1− fm)βUm,+1 + fmβWm,+1
Ur (η) = zr (η) + (1− fr (η))βmax{Ur ,+1(η),Unm,+1(η)}
+fr (η)βWr ,+1(η)
Uom = zm + (1− χf om)βUo
m,+1 + χf omβWom,+1
Unm(η) = zr (η) + (1− χf nm(η))βUn
m,+1(η) + χf nm(η)βW nm,+1(η)
with χ ∈ (0; 1) the efficiency loss in the matching process whenthe worker chooses to change occupation. The UB, zi , are indexedto the wage of the previous job i .
Back to slide model
32
Back to slide model
Routine :
w r (η) =γ
1 + τ f
(yr (η) + Γ(τ f+1, τ
w+1)
φ+1
φ(cθr (η))
)
+γ
1 + τ f
(c
qr (η)
)(1− sr )
(1− Γ(τ f+1, τ
w+1)
φ+1
φ
)
+1− γ
1− τw(zr (η) + (1− s − fr )β max{0,Un
m,+1(η)− Ur,+1(η)})
Manual (new movers) :
wnm(η) =
γ
1 + τ f
[psδAs + Γ(τ f+1, τ
w+1)
φ+1
φ
(cθnm(η)
)]
+γ
1 + τ f
(c(1− λ)
qnm(η)+
cλ
qm
)(1− sm)
[1− Γ(τ f+1, τ
w+1)
φ+1
φ
]
+1− γ
1− τw
[zr (η) + β
(λ(Un
m,+1(η)− Um,+1) + s(1− λ)(Unm,+1(η)− Uo
m,+1)
)]
with
φ =γ
1− γ(1)
Γ(τ f+1, τw+1) =
1 + τ f
1 + τ f+1
1− τw+1
1− τw(2)
32
A complex numerical algorithm
Back to slides
I A non-stationary problem : a structural change of the economy
⇒ standard methods of approximation of the dynamics around a uniquesteady state are not implementable here.
I There are several regimes
⇒ Even if we know the initial and the final steady states, the dynamicstakes into account the transitional labor reallocations (non-linear problemof occupational choice) and the MW, which can binds or not, dependingon the evolution of the economy.
I There are heterogeneous workers, and this heterogeneity matters or notdepending on the occupation of the worker. ⇒ The size of the model isvery large (more than 1500 dynamic equations).
I General equilibrium model : labor re-allocation affects relative production,hence relative price of good, hence feed-back effects on labor re-allocation
I A time-consuming process to solve this new type of problem
32
The expected interest of the analysis at the GE
I Our contributionI There is an impact of the growth on aggregate employment⇒ An unbalanced growth path leads to capitalization effects for
the favored jobs (skilled workers) and to reallocationphenomena (unskilled workers)
⇔ First general equilibrium effect.I There are consumers, and thus interaction between worker
groups through the utility function.⇒ The values of the manual jobs dependent from the abstract
and routine jobs⇔ Second general equilibrium effect.I There is a combination, specific to each country, of the
dynamics of the TBTC and LMI, affecting both the level andthe structure of the employment.
⇔ Third general equilibrium effect.
32
Non-binding scrapping-time with flexible wage
Back to slides
T = optimal
moving time
0 = initial
Tranquil time
��
Profit > 0
⇒�� > 0
But ��(�) = ��Jobs are never
destroyed
endogeneously
time
Ro
uti
ne
jo
bs
Un
em
plo
ye
dw
ork
ers Time during which
there is no hiring
��=0 because �=0
��(�)
��(�)
��(�)
32
Binding scrapping-time with rigid wage
Back to slides T = optimal Scrapping time
0 = initial
Tranquil time
Jobs are endogenously
destroyed
time
Ro
uti
ne
jo
bs
Un
em
plo
ye
dw
ork
ers
��(�)
��(�)
Time during which
there is no hiring
��=0 because profit <0
��(�)
�< ��(�)
32
Scrapping-time with rigid wage and firing costs
Back to slides
0 = initial
Tranquil time
Jobs are not
destroyed
time
Ro
uti
ne
jo
bs
Un
em
plo
ye
dw
ork
ers
Time after which there
is no hiring : �� < 0
(Optimal scrapping
Time if FC=0)
��(�)
�� � − ��
Time during which
there is no hiring
�=0 because profit <0
��(�)
��< ��(�)
��(�)
32
The interaction between moving time and scrapping time
Back to slides
T = optimal
Scrapping time
T = optimal
moving time
0 = initial
Tranquil time
Jobs are
destroyed
time
Ro
uti
ne
jo
bs
Un
em
plo
ye
dw
ork
ers
��(�)
��(�)
Profit > 0
⇒�� > 0
But �(�) = �
Time during which
there is no hiring
��=0 because ��=0
�(�)
�(�)
32
The interaction between moving time and scrapping time
J(t)
t
J(t|FC=0)
Rigid wage
0
~
>alterU
ifJ
0t
Hirings stop
because u=0
(close for the 2 cases)
J(t|FC=0 and flexible wages)
No firings
Employment goes down
at exogenous rate s
��
if FC=0
32
The interaction between moving time and scrapping time,with firing costs
-FC
J(t)
tT
If FC >0
��
if FC=0t
More incentives
to hire when FC=0
firings occur earlier
with FC = 0
Hirings stop earlier
when FC>0
J(t|FC=0) rigid wage
J(t|FC>0)
Rigid wage
0
~
>alterU
ifJ
0t
Hirings stop later
when FC=0
Employment goes down
at exogenous rate s
Employment goes down
at exogenous rate s
32
Employment reallocation in FranceBack to slides
1960 1980 20000
0.2
0.4Level
ua
usur
1960 1980 20000
0.5
1Level
na
nsnr
1960 1980 20000
0.5
1Share
na
nsnr
1960 1980 20000
0.5
1Rate
na
nsnr
1960 1980 20000
0.5
1
Moving (Umn>ur)
1960 1980 20000
0.5
1
Scraping Jr+FC<=0
1960 1980 20000
1
2x 10
-3 nmn
1960 1980 20000
0.005
0.01
0.015um
n
1960 1980 20000
0.01
0.02
0.03nm
32
Employment reallocation in the US
1960 1980 2000 20200
1
2
3
4x 10-3 Routine jobs nr(), =1
1960 1980 2000 20200
1
2
3
4
5
6x 10-3Search unemployment um
n (), =1
1960 1980 2000 20200
0.5
1
1.5
2
2.5x 10-3 Movers nm
n () , =1
32
Employment reallocation in the US
1960 1970 1980 1990 2000 2010 20200
1
2
3
4
5
6
7x 10-3 Routine jobs nr()
1960 1970 1980 1990 2000 2010 20200.26
0.28
0.3
0.32
0.34
0.36Routine jobs nr
1960 1970 1980 1990 2000 2010 20200
1
2
3
4
5
6x 10-3 Search unemployment um
n ()
1960 1970 1980 1990 2000 2010 20200
0.5
1
1.5
2
2.5
3x 10 -3 Movers nm
n ()
1960 1970 1980 1990 2000 2010 20200
0.005
0.01
0.015
0.02
0.025
0.03
0.035Manual jobs nm
33
In a frictionless labor market
Back to slides
In Autor and Dorn (2013), the impact of the Task-BiaisedTechnological Change (TBTC) is governed by two equations
Worker Mobility : ηyr = Asps Demand : ps = MRS(Cg ,Cs)
where MRS is the marginal rate of substitution between goods,and F (K , La, Lr ) = ALαa [((1− µ)Lr)σ + (µK )σ]
ασ the production
function of goods, leading to yr = F ′Lr . The mobility conditiondetermines the ability threshold η below which workers choosemanual jobs. Thus, if the elasticities of substitutions of F (·) andthe MRS(·) depend on {σ, α} and ρ respectively, then the impactof the TBTC depends only on these 3 parameters. There is nolabor supply elasticity because the supply of skilled labor is fixed inall markets.
33
In a partial equilibrium (ps is constant and exogenous), we have ina matching model :
Mobility : U(θr (η), LMI ) = U(θm, LMI )
where Hirings :
{θr (η) = ϕr (ηyr , LMI )θm = ϕm(Asps , LMI )
We deduce that the mobility between labor market segments isgoverned by :
ϕr (ηyr , LMI ) = ϕm(Asps , LMI )
As previously, the impact of the TBTC depends on {σ, α} (and ρ ifps is endogenous), but now, combined withLMI = { r , h︸︷︷︸
w r
, γ, c︸︷︷︸wNash
,MW , ω,w︸ ︷︷ ︸wage rigidity
} and thus on the labor supply
elasticity (extensive margin).
33
Assume for simplicity thatI the wage is bargained a la Hall and Milgrom (2008). In this
case, we have wr (η) = γηyr + (1− γ)(h + zr (η)) andwr (η) = γpsAs + (1− γ)(h + zm)
I There is no social programs, and the unemployment benefitsare proportional to productivity zr (η) = rηyr and zm = rpsAs ,with r the replacement ratio.
The wage becomes wr (η) = (γ + (1− γ)r)ηyr andwm = (γ + (1− γ)r)psAs .Under the assumption that yr and ps are constant (equilibriumgrowth path), mobility across labor market segments is governedby :
ηyr
[r +
βfr (η)γ(1− r)
1− β(1− s − fr (η))
]= psAs
[r +
βfmγ(1− r)
1− β(1− s − fm)
]This equation has a trivial solution : ηyr = pAm. This comes fromthe proportionality of all values function to productivity and fromthe symmetry between routine and manual functional forms(fr (·) = fm(·)). In this case, the occupational choice is governed bythe same equation as in Autor and Dorn (2013). 33
Thus, assume now that ps is constant but yr decreases at the rateg (ie. yr (t + 1) = (1− g)yr (t)). We deduce that the occupationalchoice is now given by :
ηyr
[r +
βfr (η)γ(1− r)
1− β(1− g)(1− s − fr (η))
]= pAm
[r +
βfmγ(1− r)
1− β(1− s − fm)
]The capitalization effect in the LHS, and absent in the RHS,implies that ηyr = Amps is not the equation that determines theability threshold η below which workers allocate to manual jobs.
Back to slides
33
A simple way to interpret the previous equation is to notice that itdefines η as follows :
Γ(η, g) = Υ with Γ′1(η, g) > 0 and Γ′2(η, g) < 0
When g = 0, the solution is, as previously and as in Autor andDorn (AD), ηADyr = psAm, whereas, when g > 0, η < ηAD :Search and matching reduces the magnitude of the reallocationprocess such that less workers reallocate to manual jobs. Due tosearch and matching, employment is an investment decision : timematters, and thus the capitalization of future profit flows. If profitflows are expected to decline, firms’ incentive to open vacancies isreduced. This leads workers to leave earlier the labor market of theroutine jobs than in a frictionless market. This results appears evenif wage is flexible (Nash bargaining rule) and even if there is norevenues non-indexed on wages, like social program. The gapbetween η and ηAD depends on the level of LMI, ie. in thisexample on {r , γ, c}.
33
I Need accurate data to pin down search costs
I Expected effects ?I For example, ambiguous effects for predictions on EPL
I With J2J, lower value of the firm (lower expected duration ofthe job) then, profit becomes negative sooner, hence largereffect of FC
I With J2J, more workers leave the firm before profits becomenegative, hence, there are fewer workers when profit becomesnegative, hence smaller effect of FC
33
Model parameters : values based on external information
Back to main slide
Matching c? c?a ψ? s? = s?a Υ?
0.15 2c? 0.5 0.0125 0.025Preferences β? h?s ρ ν
4% 0 0.825 0.6Technology A As σ α µ
4.5 0.95 0.78 0.6 0.5Learning δ? χ? λ?
0.9 1 0.025Wage norms ωa,US ωa,Fr ωa,Ger = ωr,Ger
0.95 0.1 0.55Adjustments gpk gLa grr ghu gMW
0.012 0.005 0.03 0.03 0.02
Blue : ”estimated” parameters
33
Model parameters : calibration dim(Φ) = dim(Ψ)
Back to main slide
The other set of parameters Φ = {Φ1,Φ2,US ,Φ2,F ,Φ2,G ,Φ3} :
Φ1 ={ρ, ν,A, σ, µ, α,As , pk(0), pk(T ), η, η, ση
}Φ2,i =
{ωa,i , hu,i (0), hu,i (T )
}i=US,F
Φ2,G ={ωa,G , ωr,G , hu,G (0) = hu,G (T ), hu,G (1995)
}The dynamics of all the exogenous variables are
x(t) = (x(0)− x(T ))e−gx t + x(T ) for t ∈ [0,T ]
This adds Φ3 = {gpk , gLa , gr , ghu , gMW } parameters, with dim(Φ) = 27.The targets of the calibration are :
Ψ =
{Na,i (0),Nr,i (0),Nm,i (0),Na,i (T ),Nr,i (T ),Nm,i (T ),Ei [Na],Ei [Nr ],Ei [Nm]
}i=US,F ,G
with dim(Ψ) = 27.
33
Benchmark case, additional graphs : US
1980 1990 20002.5
3
3.5
4
4.5p
s
1980 1990 20000.94
0.96
0.98
1
1.02y
r
1980 1990 20000
2
4
6
8y
a
1980 1990 20000
2
4
6w
a
1980 1990 20000.35
0.4
0.45
0.5
0.55Mean w
r(η )
1980 1990 20000.2
0.3
0.4
0.5Mean (w
s, w
mn (η ), w
m)
1980 1990 20000.1
0.2
0.3
0.4fa
1980 1990 20000.12
0.14
0.16
0.18
0.2Mean f
r(η )
1980 1990 20000.04
0.06
0.08
0.1Mean (f
s, f
mn (η ), f
m)
Bench. USLMI constant (1975)No TBTC
No Ls increase
Back to slides
33
Benchmark case, additional graphs : France
1980 1990 20003.5
4
4.5
5p
s
1980 1990 20000.7
0.75
0.8
0.85y
r
1980 1990 20002
4
6
8y
a
1980 1990 20000
2
4
6w
a
1980 1990 20000.4
0.5
0.6
0.7
0.8Mean w
r(η )
1980 1990 20000.4
0.6
0.8
1Mean (w
s, w
mn (η ), w
m)
1980 1990 20000.05
0.1
0.15
0.2
0.25fa
1980 1990 20000.1
0.12
0.14
0.16Mean f
r(η )
1980 1990 20000.005
0.01
0.015
0.02Mean (f
s, f
mn (η ), f
m)
Bench. FRLMI constant (1975)No TBTCNo L
s increase
Back to slides
33
Estimated shocks
Back to slides
33
Shutting down 2 trends : US
Back to slides
33
Shutting down 2 trends : France
Back to slides
33
Minimum wage (monthly, in euros)
Feui
lle2
Pag
e 1
0
200
400
600
800
1000
1200
1400
1600
France
United States
Germany (until 1990 former territory of the FRG)
Eurostat
33
Declining price of capitalBack to slides
33
Source: Source : Karabarbounis Neiman(QJE 2014).
Declining price of capitalBack to slides
1975 1980 1985 1990 1995 2000 20050.7
0.75
0.8
0.85
0.9
0.95
1
1.05
USFrance
33
Source: Investment Price (relative toconsumption price). World Development
Indicators. The relative price is normalizedto one in 1980.Source : Data from
Karabarbounis Neiman (QJE 2014).
Increase in educational attainmentBack to slides
1960 1970 1980 1990 2000 2010 20200
0.05
0.1
0.15
0.2
0.25
0.3
0.35
USFrance
33
Source: French census. Diplomeuniversitaire 1er, 2eme ou 3eme cycle,
BTS-DUT. US Census (Years of SchoolCompleted) College 4 years and more
(1) routine tasks (2) manual tasks (3) abstract tasks
(4) good market (5) occupational choice (6) service market
θr(η)
wr(η)
θs θa
wa
pspg
yg
ws
θ r(η|0)θ r(η|1)
θ s
ys
WC
JC JC JC
WC WC
AS
AD
AS
AD
���� ��
�� �� ��
��
��
��
��
η�
��
�� ��
��
33
Source: Legend : E0 (solid lines) = beforethe technological change ; From E0 to E1(dash-dot lines) = after the technological
change without General Equilibriumfeedback (no increase in ps ) ; E2 (boldlines) = after the technological change
with General Equilibrium feedback (afterincrease in ps ).
(1) routine tasks (2) manual tasks (3) abstract tasks
(4) good market (5) occupational choice (6) service market
θr(η)
wr(η)
θs θa
wa
pspg
yg
ws
θ r(η|0)θ r(η|1)
θ s
ys
WC
JC JC JC
WC WC
AS
AD
AS
AD
���� ��
�� �� ��
��
��
��
��
η�
��
��
��
��
33
Source: Legend : E0 (solid lines) = beforethe technological change ; From E0 to E1(dash-dot lines) = after the technological
change without General Equilibriumfeedback (no increase in ps ) ; E2 (boldlines) = after the technological change
with General Equilibrium feedback (afterincrease in ps ) ; bold-dashed lines = rise in
minimum wage
Back
ER EA: France
1985 1990 1995 2000 2005-2
0
2
4
6x 10-3 ER EM: France
1985 1990 1995 2000 2005-1
0
1
2
3
4x 10-3 ER UR: France
1985 1990 1995 2000 2005-0.02
0
0.02
0.04
0.06ER I: France
1985 1990 1995 2000 2005-0.02
0
0.02
0.04
0.06
ER EA: US
1980 1990 2000-0.01
-0.005
0
0.005
0.01
0.015
0.02
0.025ER EM: US
1980 1990 2000-5
0
5
10
15x 10-3 ER UR: US
1980 1990 2000-0.02
-0.01
0
0.01
0.02
0.03
0.04
0.05ER I: US
1980 1990 2000-0.02
0
0.02
0.04
0.06
34
Back
UR EA: France
1985 1990 1995 2000 2005-0.01
0
0.01
0.02
0.03
0.04UR ER: France
1985 1990 1995 2000 2005-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5UR EM: France
1985 1990 1995 2000 2005-0.02
0
0.02
0.04
0.06UR I: France
1985 1990 1995 2000 2005-0.1
0
0.1
0.2
0.3
0.4
UR EA: US
1980 1990 2000-0.02
0
0.02
0.04
0.06
0.08UR ER: US
1980 1990 2000-0.2
0
0.2
0.4
0.6UR EM: US
1980 1990 2000-0.05
0
0.05
0.1
0.15UR I: US
1980 1990 2000-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
35
Back
EA ER: France
1985 1990 1995 2000 2005-5
0
5
10x 10-3 EM ER: France
1985 1990 1995 2000 2005-5
0
5
10
15
20x 10-3 ER UR: France
1985 1990 1995 2000 2005-0.02
0
0.02
0.04
0.06I E: France
1985 1990 1995 2000 2005-0.01
0
0.01
0.02
0.03
0.04
EA ER: US
1980 1990 2000-0.01
0
0.01
0.02
0.03
0.04EM ER: US
1980 1990 2000-0.02
0
0.02
0.04
0.06ER UR: US
1980 1990 2000-0.02
0
0.02
0.04
0.06I ER: US
1980 1990 2000-0.02
0
0.02
0.04
0.06
0.08
36
Back
UA ER: France
1985 1990 1995 2000 2005-0.05
0
0.05
0.1
0.15
0.2UR ER: France
1985 1990 1995 2000 2005-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5UM ER: France
1985 1990 1995 2000 2005-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25I UR: France
1985 1990 1995 2000 2005-0.01
0
0.01
0.02
0.03
UA ER: US
1980 1990 2000-0.1
0
0.1
0.2
0.3UR ER: US
1980 1990 2000-0.2
0
0.2
0.4
0.6UM ER: US
1980 1990 2000-0.1
0
0.1
0.2
0.3I UR: US
1980 1990 2000-0.01
0
0.01
0.02
0.03
0.04
37
Worker flows
37
The story behind the disappearance of routine jobs
Documenting worker flows in France and in the US
I Survey data : US CPS (monthly, Jan 1976-July 2016) andFrench LFS (annual, 1983-2014)
I use current or most recent occupation to categorizeindividuals into task groups : Abstract, Routine, Manual
I Compute each period transition rates between 7 states :Employed (Abstract, Routine, Manual), Unemployed(Abstract, Routine, Manual), Not in Labor Force
I Trends (HP-filter), sample stops before the 2008 crisis
ER → EA,ER → EM, ER → UR, ER → I Figure
UR → EA,UR → ER, UR → EM, UR → I Figure
EA→ ER,EM → ER, ER → UR, I → ER Figure
UA→ ER,UR → ER, UM → ER, I → UR Figure
38
Table – Transition matrix(t) : 7×7 matrix of transition rates
EA(t+1) ER(t+1) EM(t+1) UA(t+1) UR(t+1) UM(t+1) I(t+1)
EA(t)
ER(t)
EM(t)
UA(t)
UR(t)
UM(t)
I(t)
stocks(t+1)=matrix(t) × stocks(t).stocks are a vector of size (7,1) each period
39
The employment shares : a measure of polarization
1970 1980 1990 2000 20100.15
0.2
0.25
0.3EA/E
1970 1980 1990 2000 20100.25
0.3
0.35
0.4
1970 1980 1990 2000 20100.6
0.65
0.7
0.75
0.8ER/E
1970 1980 1990 2000 20100.45
0.5
0.55
0.6
0.65
1970 1980 1990 2000 20100.1
0.12
0.14
0.16EM/E
1970 1980 1990 2000 20100.12
0.14
0.16
0.18
FranceUS
I Shares of abstract and services jobs increased in bothcountries.
I Routine jobs : a great similarity between France and theUnited States ⇔ a decrease of about 10 points.
40
The employment shares : a measure of polarization
Common decline in the share of routine jobs in total employmentbut for different reasons
41
Employment levels :
Routine per-capita employment is different across countries.
1970 1980 1990 2000 20100
0.1
0.2EA
1970 1980 1990 2000 20100.15
0.2
0.25
1970 1980 1990 2000 20100.2
0.4
0.6ER
1970 1980 1990 2000 20100.3
0.35
0.4
1970 1980 1990 2000 20100.055
0.06
0.065
0.07
0.075EM
1970 1980 1990 2000 20100.08
0.09
0.1
0.11
0.12
1970 1980 1990 2000 20100.4
0.5
0.6
0.7E
1970 1980 1990 2000 20100.6
0.65
0.7
0.75
FranceUS
42
Employment levels :
1970 1975 1980 1985 1990 1995 2000 2005 20100
0.1
0.2EA
1970 1975 1980 1985 1990 1995 2000 2005 20100.15
0.2
0.25
1970 1975 1980 1985 1990 1995 2000 2005 20100.2
0.4
0.6ER
1970 1975 1980 1985 1990 1995 2000 2005 20100.3
0.35
0.4
1970 1975 1980 1985 1990 1995 2000 2005 20100.055
0.06
0.065
0.07
0.075EM
1970 1975 1980 1985 1990 1995 2000 2005 20100.08
0.09
0.1
0.11
0.12
1970 1975 1980 1985 1990 1995 2000 2005 20100.4
0.5
0.6
0.7E
1970 1975 1980 1985 1990 1995 2000 2005 20100.6
0.65
0.7
0.75FranceUS
Constant Routine jobs
↑ total Edue to Abstract and Manual jobs
↓ Rou ne jobs
↓ Total
43
Employment levels :
1970 1975 1980 1985 1990 1995 2000 2005 20100
0.1
0.2EA
1970 1975 1980 1985 1990 1995 2000 2005 20100.15
0.2
0.25
1970 1975 1980 1985 1990 1995 2000 2005 20100.2
0.4
0.6ER
1970 1975 1980 1985 1990 1995 2000 2005 20100.3
0.35
0.4
1970 1975 1980 1985 1990 1995 2000 2005 20100.055
0.06
0.065
0.07
0.075EM
1970 1975 1980 1985 1990 1995 2000 2005 20100.08
0.09
0.1
0.11
0.12
1970 1975 1980 1985 1990 1995 2000 2005 20100.4
0.5
0.6
0.7E
1970 1975 1980 1985 1990 1995 2000 2005 20100.6
0.65
0.7
0.75FranceUS
↓ Rou ne jobs
↑ Rou ne jobs
↑ Total
44
The employment shares : a measure of polarization
Common decline in the share of routine jobs in total employmentbut for different reasons
Any analysis based on employment shares alone provides a partialpicture of job polarization
45
Dynamics of Routine Employment
The decline in routine jobs in France is the main driver of theemployment gap with respect to the US
1980 1990 2000 20100.45
0.5
0.55
0.6
Ec|AFR
data France
France with EAUS
1980 1990 2000 20100.45
0.5
0.55
0.6
Ec|RFR
data France
France with ERUS
1980 1990 2000 20100.45
0.5
0.55
0.6
Ec|MFR
data France
France with EMUS
1980 1990 2000 20100.45
0.5
0.55
0.6
0.65
EcFR
data France
France with ERUS
France with {ERUS,EAUS}
France with {ERUS,EAUS, EAUS}
46
Counterfactuals
I Which changes in transition rates are key in accounting forthe evolution of per capita routine employment ?
I Looking at the evolution of transition rates is not enough astransition rates interact with stocks
I Counterfactual experimentsI fix some transition rates at their level at the beginning of the
sample ⇒ get counterfactual transition ratesI Using the initial stocks, iterate forward using the
counterfactual transition rates :stocks(t+1)=matrix(t)×stock(t)
I predict the counterfactual routine per-capita employmentI compute the gap between the observed and counterfactual
evolutions of routine employmentI how much of the fall in routine employment would have been
prevented if particular transition rates had remained at thelevels observed prior to the onset of job polarization ?
I as in Cortes, Jaimovich, Nekarda and Siu (2015)
47
Counterfactuals
Ins and Outs of routine employment
I Inactivity : Labor market participation
I Employment : Job-to-job, occupational mobility
I Unemployment : Job finding, job separation
48
INs-OUTs of Routine jobs from Inactivity
1980 1985 1990 1995 2000 2005 2010
0.3
0.32
0.34
0.36
0.38
0.4France: Counterfactual Routine per-capita Employment
I to {E,U}{E,U} to I{E,U} <-> IData
1980 1985 1990 1995 2000 2005 2010-0.04
-0.03
-0.02
-0.01
0
0.01France: Gap = counterfactual - data
I to {E,U}{E,U} to I{E,U} <-> I
1975 1980 1985 1990 1995 2000 2005 2010
0.32
0.34
0.36
0.38
0.4US: Counterfactual Routine per-capita Employment
I to {E,U}{E,U} to I{E,U} <-> IData
1975 1980 1985 1990 1995 2000 2005 2010-0.02
-0.01
0
0.01
0.02
0.03US: Gap = counterfactual - data
I to {E,U}{E,U} to I{E,U} <-> I
out
out
In &out
In &out
I France : +4pp of ER generated by a decline in the outflows toinactivity (pension reforms ?)
I US : after 1995, -2.5pp in ER due to an increase in theoutflows to inactivity, consistent with Cortes et al (2015)
49
Occupational mobility
1980 1985 1990 1995 2000 2005 2010
0.3
0.32
0.34
0.36
0.38
0.4France: Counterfactual Routine per-capita Employment
ER to EAER to EMER -> {EA,EM}Benchmark
1980 1985 1990 1995 2000 2005 2010-2.5
-2
-1.5
-1
-0.5
0
0.5
1x 10-3 France: Gap = counterfactual - data
ER to EAER to EMER -> {EA,EM}
1975 1980 1985 1990 1995 2000 2005 20100.32
0.34
0.36
0.38
0.4
0.42
0.44US: Counterfactual Routine per-capita Employment
ER to EAER to EMER -> {EA,EM}Benchmark
1975 1980 1985 1990 1995 2000 2005 2010-0.02
-0.01
0
0.01
0.02
0.03US: Gap = counterfactual - data
ER to EAER to EMER -> {EA,EM}
I France : @ changes in mobility across occupations
I US : -2pp in ER due to a rise in occupational mobility,whether downward or upward
50
INs-OUTs of Routine jobs from Unemployment
1980 1985 1990 1995 2000 2005 2010
0.3
0.32
0.34
0.36
0.38
0.4France: Counterfactual Routine per-capita Employment
ER to URU to ER ER <-> UR or UData
1980 1985 1990 1995 2000 2005 2010-0.02
-0.015
-0.01
-0.005
0
0.005
0.01
0.015France: Gap = counterfactual - data
ER to URU to ER ER <-> UR or U
1975 1980 1985 1990 1995 2000 2005 2010
0.32
0.34
0.36
0.38
0.4US: Counterfactual Routine per-capita Employment
ER to URU to ER ER <-> UR or UData
1975 1980 1985 1990 1995 2000 2005 2010-0.02
-0.015
-0.01
-0.005
0
0.005
0.01US: Gap = counterfactual - data
ER to URU to ER ER <-> UR or U
JFRJSR
I in France, in the mid-1990s, - 1 pp of ER due to high JSR.After the mid-1990s, the increase in ER is driven by higherJFR.
I The In-Out flows from U does not explain ER in the US.
51
Summary
France USRoutine N Nr down Nr flat
then up after 1995 then down after early 1990sInactivity less outflows to inactivity more outflows to inactivity
especially after mid-1990sJob-to-Job more outflows to others jobsUnemployment high JSR before mid-1990s
high JFR after mid-1990s
Consistent with the view that, in France,
I Lower payroll taxes for low-paid jobs since the late 1990s
I → rebound in routine employment = Protection of routinejobs
I Hence, workers are not enticed to look for a job elsewhere(low job-to-job)
52
Research Agenda
Model with additional features
I Endogenous participation
I Job-2-JobI Institutional differences :
I France : payroll tax subsidy on low-wage workers and pensionreforms increase retirement age ⇒ life-cycle features in themodel
I US : identifying cross-country difference in cost of job-2-jobmobility
53
Occupational choices, LMIs and General Equilibrium effects
In Autor and Dorn (2013) : No labor market frictions, Mobilitychoice based on wage comparison :
I wage routine wr = ws wage service
I ability threshold η such that η > η = routine
Mobility : ηyr = AspsDemand : ps = MRS(Cg ,Cs)
}⇒ η = φw (σ, α, ρ)
I Good production function : σ, α (technological parameters),↓ pk ⇒↓ cost of routine tasks and ↑ capital ⇒ ↑ supply ofgoods ⇒ ↑ demand for goods
I ρ (consumer preference, must favor variety) : so that demandfor service ↑
General equilibrium effect through ps : ↑ ps is a signal that routineworkers shall switch to manual jobs
54
Occupational choices, LMIs and GE effects
Our paper : ”search unemployment”
Mobility : U(θr (η), LMI ) = U(θm, LMI )
Demand :
{θr (η) = ϕr (ηyr , LMI )θm = ϕm(Asps , LMI )
⇒ η = φSaM({σ, α, ρ}, LMI )
where : LMI = {rr , h︸︷︷︸w r
, γ, c︸︷︷︸wNash
, MW︸︷︷︸wage rigidity
, taxes}
General equilibrium effect through psLMI on both sides of equations but does not go away because ofcapitalization effect (as long as divergent evolution of productivityacross sectors) Example
”Rest unemployment” : θm = 0. Reallocations are stalled.Solution method in a non-stationary, non-linear general-equilibrium model with heterogenous agents
Example : France
55
Welfare analysis :Winners and losers of the US
and French polarizationB. Obama, The Economist, October 2016.”What is happening in the American political system ?How has a country that has benefited - perhaps more than anyother - from [...] technological innovation suddenly developed astrain of [...] anti-innovation protectionism ?Why have some [...] embraced a crude populism ?”
55
Measuring welfare
I Value functions W ,U that includeI income flowI future employment opportunities
I Average measures based of number of workers in eachcategory (employment status, task, ability η)
I divided by consumer price index (which increases due to risein relative price of services)
56
Table – Winners and losers of the US and French polarization
US FranceCommon features : Average welfare ↑
Manual and abstract workers arethe winners of JP
Difference :The fate of The US sacrificed ”stayers” France favored ”stayers”routine workers and favored to the detrimentlinked ”movers” of ”movers”to LMI changes to ↑ employment
57
Abstract and manual workers are the winners of JP
1975 1980 1985 1990 1995 2000 2005100
110
120
130
140
150
160
170US - Employees - Abstract
years
1975
= 1
00
1975 1980 1985 1990 1995 2000 2005100
120
140
160
180
200US - Employeees - Manual
years
1975
& o
ld m
over
= 1
00
New moversOld moverManualAverage
1975 1980 1985 1990 1995 2000 2005100
110
120
130
140
150
160
170US - Unemployed - Abstract
years
1975
= 1
00
1975 1980 1985 1990 1995 2000 2005100
120
140
160
180
200
220US - Unemployed - Manual
years
1975
& o
ld m
over
= 1
00
Index base 100 = 1975.58
Abstract and manual workers are the winners of JP
1980 1990 2000100
110
120
130
140
150France - Employees - Abstract
years
1975
= 1
00
1980 1990 2000100
150
200
250France - Employeees - Manual
years
1975
& o
ld m
over
= 1
00
New moversOld moverManualAverage
1980 1990 2000100
110
120
130
140
150France - Unemployed - Abstract
years
1975
= 1
00
1980 1990 200050
100
150
200
250
300France - Unemployed - Manual
years
1975
& o
ld m
over
= 1
00
Index base 100 = 1975.59
Figure – US : routine workers lose along the transition
1975 1980 1985 1990 1995 2000 2005
100
150
200
250
300
350
US - Employees - Routine
years
1975
& (
=1)
= 1
00
Still inOutAverage
1975 1980 1985 1990 1995 2000 2005
100
150
200
250
300
350
US - Unemployed - Routine
years19
75 &
(
=1)
= 1
00
Index base 100 = 1975 for η, and Index base 100 ∗Welfare(η)/Welfare(η) = 1975 for η > η. “Still in” : inroutine pool, the highest welfare line is the routine worker with the highest productivity η, “Out“ : displacedroutine worker (with the lowest productivity η) who switch to manual job. 60
France : divided middle class
1975 1980 1985 1990 1995 2000 2005
100
150
200
250
300
350
400France - Employees - Routine
years
1975
& (
=1)
= 1
00
Still inOutAverage
1975 1980 1985 1990 1995 2000 2005
100
150
200
250
300
350
400France - Unemployed - Routine
years19
75 &
(
=1)
= 1
00
Index base 100 = 1975 for η, and Index base 100 ∗Welfare(η)/Welfare(η) = 1975 for η > η. “Still in” : inroutine pool, the highest welfare line is the routine worker with the highest productivity η, “Out“ : displacedroutine worker (with the lowest productivity η) who switch to manual job. 61
Abstract and manual welfare : counterfactuals
1980 1990 2000
years
-150
-100
-50
0
50
%US - Abstract Employees
No TBCTConstant LMIConstant La
1980 1990 2000
years
-150
-100
-50
0
50US - Abstract Unemployed
1980 1990 2000
years
-40
-30
-20
-10
0
10
20
%
US - Manual Employees
1980 1990 2000
years
-40
-30
-20
-10
0
10
20US - Manual Unemployed
1980 1990 2000
years
-100
-80
-60
-40
-20
0
20France - Abstract Employees
1980 1990 2000
years
-150
-100
-50
0
50France - Abstract Unemployed
1980 1990 2000
years
-40
-35
-30
-25
-20
-15
-10France - Manual Employees
1980 1990 2000
years
-25
-20
-15
-10
-5
0France - Manual Unemployed
“%“ : welfare gap (in percentage) between benchmark simulation and counterfactual experiment. “Constant LMI“ :LMI set at 1975 value. “No TBTC“ : price of capital pk is constant. “Constant La“ : La constant, set at 1975 value.
62
Routine welfare : US counterfactuals
020
4060
80100
1980
1990
2000
0
5
10
15
20
abilities
US - Employees - LMI
years
%
“%“ : welfare gap (in percentage) between benchmark simulation and counterfactual experiment. Gap>0 : routineworkers are happier with conterfactual. 63
Routine welfare : US counterfactuals
64
Routine welfare : France counterfactuals
020
4060
80100
1980
1990
2000
-15
-10
-5
0
5
abilities
France - Employees - LMI
years
%
“%“ : welfare gap (in percentage) between benchmark simulation and counterfactual experiment. Gap>0 : routineworkers are happier with conterfactual. 65
Routine welfare : French counterfactuals
“%“ : welfare gap (in percentage) between benchmark simulation and counterfactual experiment. “Constant LMI“ :LMI set at 1975 value. “No TBTC“ : price of capital pk is constant. “Constant La“ : La constant, set at 1975 value. 66
Conclusion
Job polarization using US and French data.
I Dynamics of employment shares are very similar acrosscountries.
I but major differences in the dynamics of routine employmentlevelsI US routine employment level actually increase until the early
1990s, then started falling.I The evolution of French routine employment went in opposite
directions to that of the US economy.
67