A Tale of Two Countries: A Story of the French and US ... · 1 1.05 US France 4 Source: Investment Price (relative to consumption price). World Development Indicators. The relative

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


manual tasks


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


at exogenous rate s


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


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


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


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


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


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


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


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


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

Documents

A Tale of Two Countries: A Story of the French and US ... · 1 1.05 US France 4 Source: Investment Price (relative to consumption price). World Development Indicators. The relative