Labor Market Conflict and the Decline of the Rust Belt Simeon Alder1

David Lagakos2 1 Notre

Dame

2 UCSD

and NBER

3 UCLA

and NBER

March 25, 2017

Lee Ohanian3

The Rust Belt

. . . or, as of November 8, 2016 . . .

Rust Belt Employment Share 0.35 0.40 0.45 0.50

0.55

The Decline of the Rust Belt’s Share of U.S. Employment

Manufacturing, ex Sun Belt

Manufacturing

0.30

Aggregate

1950

1960

1970

1980

1990

2000

Four Facts About Rust Belt Since WW II

1. Labor markets were rife with conflict / threats of strikes

2. Rust Belt wages substantially higher than average

3. Weak productivity growth in Rust Belt industries

4. Reversal of 1980s: Rust Belt decline slowed, less conflict, wage premia declined & productivity growth increased

Our Theory of the Rust Belt’s Decline I

Labor market conflict lead to a hold-up problem

I

This depressed investment rates and productivity growth

I

⇒ Production moves from Rust Belt to Rest of U.S.

I

Key: slow secular decline in Rust Belt’s employment share

I

More efficient labor relations in 1980s ⇒ higher investment, stabilization of Rust Belt’s employment share

Our Theory of the Rust Belt’s Decline I

Labor market conflict lead to a hold-up problem

I

This depressed investment rates and productivity growth

I

⇒ Production moves from Rust Belt to Rest of U.S.

I

Key: slow secular decline in Rust Belt’s employment share

I

More efficient labor relations in 1980s ⇒ higher investment, stabilization of Rust Belt’s employment share

I

Alternative hypothesis: foreign competition hurt Rust Belt more than rest of U.S. – Timing and magnitudes wrong

Related Literature I Competition & productivity: Acemoglu, Akcigit (2011), Aghion, Akcigit,

Howitt (2014), Aghion, Bloom, Blundell, Griffith, Howitt (2005), Bloom, Draca, Van Reenan (2016), Cole, Ohanian (2004), Collard-Wexler, De Loecker (2015), Dunn, Klimek, Schmitz (2010), Herrendorf, Texeira (2011), Holmes (1998), Holmes, Schmitz (2010), Schmitz (2005), Syverson (2011), Parente, Prescott (1999), Pavcnik (2002) . . . I Unions and economic performance: Acikgoz, Kaymak (2012), Borjas,

Ramey (1995), Bridgman (2015), Dinlersoz, Greenwood (2012), Holmes (1998), Krueger, Mas (2004), Taschereau-Dumouchel (2015) . . . I Rust Belt / U.S. regional dynamics: Autor, Dorn, Hanson (2013,

2014), Blanchard, Katz (1992), Desmet and Rossi-Hansberg (2009, 2014), Duranton and Puga (2009), Feyrer, Sacerdote, Stern (2007), Glaeser, Ponzetto (2007), Kahn (2009), Yoon (2017) . . .

This Talk

1. Four facts about Rust Belt’s decline 2. Model 3. Quantitative analysis 4. Supporting evidence from U.S. cities and industries

Four Facts About Rust Belt’s Decline

Fact 1: Labor Market Conflict in the Rust Belt

I

Rust Belt’s main industries dominated by powerful unions (United Auto Workers, United Steelworkers...)

I

National Academy of Sciences, 1982: “adversarial and bitter relationship between labor and management”

I

Strike threat dominated post-war labor relations

I

I

steel strikes: every 3-4 years from 1946 through 1970s

I

similar in automobiles, rubber tires, equipment manufacturing

U.S. Bureau of Labor Statistics: “most concentrated period of labor-management strife in the country’s history”

Labor Market Conflict and Inefficient Rent Sharing

I

Literature: labor market conflict ⇒ inefficient rent sharing

I

Producers and unions agree with this conclusion

I

I

”Management refuses to show us the books.”

I

”We went on strike because we didn’t trust what they told us.”

I

”When profitable, we accept union demands & recoup with higher prices. Strikes risk ruining the company.”

Agreement in literature that inefficient rent sharing depressed investment & productivity growth

Labor Market Conflict Declines in 1980s #!!"

!"#$%&'()'*(&+',-(../0%1''

'#!" '!!" &#!" &!!" %#!" %!!" $#!" $!!" #!" !" $(')" $(#%" $(#)" $(*%" $(*)" $()%" $())" $(+%" $(+)" $((%" $(()" %!!%" %!!)" %!$%"

1.00

1.05

Relative Wage 1.10

1.15

1.20

Fact 2: Rust Belt Wages High

1950

1960

1970 Simple Ratio

1980

1990 With Controls

2000

Fact 2: Rust Belt Wages High

I

Not higher cost of living; very similar in BLS study of 1967

I

Unlikely to be just better workers...

I

Wage loss from displaced workers shows greater losses after plant closing for Rust Belt workers than others

I

Historical studies claim Rust Belt workers rationed jobs and earned rents

Fact 3: Rust Belt Productivity Growth Low Labor Productivity Growth in Rust Belt Industries Annualized Growth Rate, % 1958-1985

1985-1997

1958-1997

Blast furnaces, steelworks, mills

0.9

7.6

2.8

Engines turbines

2.3

2.9

2.5

Iron and steel foundries

1.5

2.3

1.7

Metal forgings/stampings

1.5

2.8

1.9

Metalworking machinery

0.9

3.5

1.6

Motor vehicles/equipment

2.5

3.8

2.9

Photographic equipment/supplies

4.7

5.1

4.9

Railroad locomotives/equipment

1.6

3.1

2.0

Screw machine products

1.2

1.1

1.2

Rust Belt weighted average

2.0

4.2

2.6

Manufacturing weighted average

2.6

3.2

2.8

Rust Belt was Technological Laggard

I

I

Autos, steel, rubber did not adopt latest technologies: I

National Academy of Sciences: producers did not adopt long-available technologies (e.g. basic oxygen furnace, continuous caster, electric arc furnace, . . . )

I

McKinsey productivity study on autos: slow adoption of “lean production” in autos

I

Literature comparing productivity to other countries: US producers were slow to roll out new products (e.g. radial tires, fuel-efficient engines, . . . )

Frequently cite bad labor relations as important factor

Model

Households

I

Unit measure of individuals

I

Labor supply inelastic

I

Preferences

∞ X t=0

δ t Ct

Final Good I

Final good produced from continuum of intermediates indexed by i (industry), j (variety), and origin: Z

1

yt (i)

Yt =

σ−1 σ

σ  σ−1 di

0

Z yt (i) =

yt (i, j) 0

where

∗

1

ρ−1 ρ

Z dj +

1

yt∗ (i, ˜)

ρ−1 ρ

ρ  ρ−1 d˜  ,

0

denotes varieties produced abroad

I

As in Atkeson, Burstein (2008); Edmond, Midrigan, Xu (2015)

I

Intermediates are gross substitutes, and ρ > σ > 1

Intermediate Goods and Two Regions

I

Industries i ∈ [0, λ) located in Rust Belt (R)

I

Industries i ∈ [λ, 1] located in Rest-of-Country (S)

I

Regions differ by extent of competition in labor markets (captured by time-varying union bargaining power βt )

Intermediate Firms

1. Produce output (variety j in industry i) using labor: yt = zt · `t
2. Invest C x, z, Z units of the final good, increase productivity by 100 · x percent next period: zt+1 = zt (1 + xt )

International Trade and Foreign Sector

I

Intermediate varieties are tradable between US and Rest-of-World

I

International trade in intermediates is subject to time-varying (symmetric) iceberg costs τt ≥ 1

I

Final goods and labor are non-tradable internationally

I

Trade is balanced period-by-period

I

Foreign producers have productivity zt (i)∗

I

Productivity grows exogenously: zt+1 (i)∗ = zt∗ (i)(1 + χ)

Labor Markets

I

Labor market competitive in Rest-of-Country

I

Rust Belt labor market governed by union

I

Individual worker’s union status denoted by υ ∈ {0, 1} and unionization rate in economy denoted by U ∈ [0, 1]

I

Workers retire with probability ζ each period (regardless of union status)

I

Access to Rust Belt jobs restricted to union members

I

Non-members can apply for vacant Rust Belt jobs and new union members are selected randomly from applicant pool

Union and Bargaining

I

Union bargains with Rust Belt producers over profits

I

Nash bargaining with time-varying union weight βt

Exogenous State Variables

I

Two exogenous state variables: 1. union bargaining power β ≥ 0 2. iceberg trade cost τ ≥ 1

I

Follow a Markov process:

(βH , τH ) (βL , τL )

(βH , τH )

(βL , τL )

1− 0

 1

Aggregate Endogenous State Variables

I

Aggregate unionization rate U with law of motion:   U 0 = (1 − ζ) U + M F (1 − U )

I

 1  1 Set of productivities Z ≡ z(i, j) i,j=0 ∪ z ∗ (i, j) i,j=0

Idiosyncratic Endogenous State Variables

I

Intermediate firm’s productivity z(i, j) with law of motion: z 0 (i, j) = z(i, j) 1 + x(i, j)

I




Worker’s union status υ with following law of motion: I

I

If υ = 1, then υ 0 = 1 with probability 1 − ζ ( 1 with probability F if R, If υ = 0, then υ 0 = 0 with probability 1 − F if R, or if S

Intermediate Firms’ Static Problem (Production) n

o Π Z, U, z; β, τ = max∗ p · y + p∗ · y ∗ − n + n∗ n, n

subject to y =z·n y = P (Z, U ; β, τ )σ−1 P (i; Z, U ; β, τ )ρ−σ | {z } {z } | agg. price index

ind. price index

−ρ

· X(Z, U ; β, τ ) p | {z } agg. expenditures

Intermediate Firms’ Static Problem (Production) n

o Π Z, U, z; β, τ = max∗ p · y + p∗ · y ∗ − n + n∗ n, n

subject to y ∗ = z · n∗ y(i, j)∗ = P ∗ (Z, U ; β, τ )σ−1 P ∗ (i; Z, U ; β, τ )ρ−σ | {z } {z } | agg. price index

∗

ind. price index

∗ −ρ

· X (Z, U ; β, τ )(p ) | {z } agg. expenditures

Profit-maximizing price is Dixit-Stiglitz markup: p = p∗ =

ρ −1 ρ−1 z

Intermediate Firms’ Dynamic Problem (Innovation) In the Rest-of-Country: V s (Z, U, zs ; β, τ ) = maxxs >0

n Πs (Z, U, zs ; β, τ ) −P (Z, U ; β, τ ) · C(xs , zs , Z) h io +δE V s (Z 0 , U 0 , zs0 ; β 0 , τ 0 ) ,

In the Rust Belt: V r (Z, U, zr ; β, τ ) = maxxr >0 {(1 − β)Πr (Z, U, zr ; β, τ ) −P (Z, U ; β, τ ) · C(xr , zr , Z) h io +δE V r (Z 0 , U 0 , zr0 ; β 0 , τ 0 ) , → Hold-up problem when β > 0

Worker’s Problem

 W (Z, U, M, υ; β, τ ) = max W r (Z, U, M, υ; β, τ ), W s (Z, U, υ; β, τ ) Value of non-union worker in Rust Belt: n W r (Z, U, M, 0; β, τ ) = F (Z, U, M ; β, τ ) w + R(Z, U ; β, τ )   o +δ (1 − ζ)E W (Z 0 , U 0 , M 0 , 1; β 0 , τ 0 )
+ 1 − F (Z, U, M ; β, τ ) n  o × w−u ¯ + δE W (Z 0 , U 0 , M 0 , 0; β 0 , τ 0 ) , where u ¯ ≥ 0.

Worker’s Problem (cont’d)

Value of union worker in Rust Belt: W r (Z, U, ·, 1; β, τ ) = w + R(Z, U ; β, τ )   +δ(1 − ζ)E W (Z 0 , U 0 , M 0 , 1; β 0 , τ 0 )

Value of any worker in the Sun Belt:   W s (Z, U, υ; β, τ ) = w + δ(1 − ζ)E W (Z 0 , U 0 , υ; β 0 , τ 0 )

Balanced Trade

I

Domestic labor is num´eraire

I

Trade balance requirement pins down foreign wage w∗ (τ, Z)

Equilibrium Properties of Model

I

Union workers strictly prefer to stay in Rust Belt

I

Non-union workers indifferent between lower wages in ROC and costly queuing for union job

I

While in (βH , τH ) state, Rust Belt invests less than ROC

I

zr rises slower than zs , so relative price of Rust Belt goods rise

I

Rust Belt share of employment shrinks through union attrition as relative demand for Rust Belt goods fall

Quantitative Analysis

Quantitative Analysis

I

How large is model’s predicted decline in Rust Belt employment share?

I

Discipline quantitative exercise by extent of competition: 1. from foreign producers (regional trade shares, 1950-2000) 2. in labor markets (estimated wage premiums, 1950-2000)

I

Import shares are low and wage premiums high initially, then move in opposite directions starting in 1980s

Calibration

I

Calibrate to manufacturing sector

I

Model and data period = 5 years

I

Set δ = 0.965 to match 4% annual interest rate

I

Set σ = 2.7 (Broda and Weinstein, 2006)

I

Set ρ = 4 to match 1/3 markup

I

Normalize initial domestic productivities to 1

State of Competition

I I

State of competition denoted by (βt , τt )  T Ex post path (βt , τt )1950+5t t=0 evolves deterministically with transition from H to L in 1985

Cost of Innovation

Cost function in units of domestic final good for i ∈ {R, S}:
z(i, j)ρ−1 C x(i, j), z(i, j), Z = α · x(i, j)γ · D(Z)

,

where α > 0, γ > 1, 2−ρ  h i 1−σ i 1−σ  1−σ h ρ−1 ρ−1 ∗ ρ−1 1−ρ ∗ ρ−1 1−ρ D(Z) = λ ZR + ZR + (1 − λ) ZS + ZS

Why? Delivers balanced growth and stationary dynamic program

Calibration: Parameters

I

(τH , τL ) – iceberg trade costs

I

∗ zR,1950 – “foreign Rust Belt” productivity in 1950

I

(βH , βL ) – union’s bargaining weight

I

λ – share of varieties produced by Rust Belt

I

α – linear (scale) parameter of cost function

I

γ – curvature parameter of cost function

Calibration: Targets

I

(τH , τL ) – aggregate import shares: 4% (pre-1985) and 9% (post-1985)

I

∗ zR,1950 – Rust Belt import share: 8% (pre-1985)

I

(βH , βL ) – Wage premium: 12% (pre-1985), 4% (post-1985)

I

λ – Initial Rust Belt employment share of 51.3%

I

α – 1.8% aggregate TFP growth (1950-2000)

I

γ – 8.5% Investment-to-GDP ratio (1950-2000) (value added share of R&D, advertising, and intangible expenditures)

Rust Belt Employment Share 0.35 0.40 0.45 0.50

0.55

Rust Belt Employment Share in Model and Data

Model

0.30

Data

1950

1960

1970

1980

1990

2000

30

40

Import Shares in Model and Data

Import Share 20

Steel & Autos

Rust Belt, model

10

Aggregate, model

0

Aggregate, data

1950

1960

1970

1980 year

1990

2000

Annualized Productivity Growth Rate .5 1 1.5 2 2.5

3

Productivity Growth Rate in Model

Rest of Country

0

Rust Belt

1950

1960

1970

1980

1990

2000

0.55

Decomposing Rust Belt’s Decline

Rust Belt Employment Share 0.35 0.40 0.45 0.50

No union hold−up

0.30

No fall in trade costs

1950

1960

1970

1980

1990

2000

Sensitivity of Quantitative Results

Alternative Specification

Rust Belt EmploymentShare Decline (%)

Lower outer elasticity, σ = 2.0

8.7

Benchmark, σ = 2.7

9.8

Higher outer elasticity, σ = 3.0

11.6

Lower inner elasticity, ρ = 3.5

9.5

Higher inner elasticity, ρ = 4.5

9.4

Investment/GDP of 8 percent

9.6

Investment/GDP of 9 percent

10.3

Supporting Evidence

‘Wage Premia” and Employment Growth at City Level

I

Use U.S. census micro data for 1950 - 2000

I

Run Mincer-type regression for 1950 log wi,m = α · SCHi,m +

4 X

j βj · EXPi,m +

j=1

M X m=1

I

Wage premium for MSA m: πm

I

Crude but useful proxy for labor union power

Dm · πm + εi,m

6.0

Employment Growth Low Where Premiums High ORL PHX

Employment Growth 1950 to 2000 0.0 2.0 4.0

AUS

RAL

GRB SJO ATL SDG SLC SAC DAL ABQ HOU DEN NSV COM WAS CHR NOR LTL SAT MIA BAT KAL SEA GRR TAC JCK JSV MAD STO FRE SCR MININD AUG MOB TUL PTV CHO SAG FTW LUB ELP OKC RCM ATC GRN HAR CLB WIC MNT LNG WAC ASHYOR SHV SYR LAN KNX LIN LIN WLM CDR RCF SPO KAN HAM GALLOS RCH MEMCOR OMA DES DAY SXF LUI BRK AMR SAV POR NOL ROA REA ALT EVABEA JCS ALL BAL TER STL CIN CLM AKR PEO MIL TOP BIR TOL PROMUN JOH CAN RAC ALB CLE DET UTI ERE DEC PHIBNG TRE CHI BOS WIF SPH SOB SIXDAV DUL HRT PIT SPR PUE SFR WAR BUF MAN BRG NHANYO FLI WOR TAM

CHT

GNA

−0.4

−0.2

0.0 Wage Premium in 1950

0.2

0.4

Industry Unionization Rates and Employment Growth

I

Use Current Population Survey (CPS) micro data

I

Collapse to industry level

I

Compute difference between unionization rate in Rust Belt and rest of U.S. for each industry

I

... same for annualized employment growth from 1950 to 2000

Industry Unionization Rates and Employment Growth

1 0 −1 −2 −3 −4 −5

Employment Growth Rate: Rust Belt − Rest of US

2

Employment Growth and Unionization Rates: Rust Belt − Rest of US

−0.1

0.0

0.1

0.2

Unionization Rate: Rust Belt − Rest of US

0.3

0.4

Conclusion I

No region of U.S. declined more than the Rust Belt since 1950

I

Labor market conflict and resulting inefficient lack of investment plays primary role

I

“Foreign competition” story gets timing, magnitudes wrong

I

Supported by evidence from cross-section of U.S. cities and industries

Extra Slides

Inefficient Rent Sharing and Low Productivity Growth Literature emphasizes that labor conflicts contribute to low investment and productivity: I

National Academy of Sciences: collective bargaining contracts that grew from adversarial and bitter relationships negatively affected technology and innovation

I

Census Bureau: Bad labor relations negatively affect productivity and costs

I

Literature comparing US to Japan: High U.S. costs reflect bad labor relations

I

President of Germany’s auto union: “German workers don’t strike. American unions are destructive because of a hostile environment. American unions can’t be constructive if their as*** are always getting kicked.”

Additional Evidence

I

Estimates of R&D intensity by industry from 1970s

I

TFP growth: Rust Belt vs. Japan

I

Adoption rates for key technologies

Post-1980 Growth

Conclusion

TFP

Adoption

R&D

R&D by Industry: Evidence I

Report of U.S. Office of Technology Assessment (1980)

I

Average manufacturing industry: R&D to Sales of 2.5%

I

Industries with highest ratio:

I

I

Communications equipment: 15.2%

I

Aircraft and parts: 12.4%

I

Office and computing equipment: 11.6%

Rust Belt industries: I

Autos: 2.1%

I

Rubber and Plastics: 1.2%

I

Steel: 0.4% Additional Evidence

Productivity Growth: United States versus Japan

I

Steel (Lieberman and Johnson, 1999) US: TFP growth < 1 percent per year 1950 to 1970 Japan: TFP doubled over same period

I

Autos (Fuss and Waverman, 1991) US: 1.6 percent per year in 1970s Japan: 4.3 percent per year in 1970s Additional Evidence

Technology Adoption: Evidence

I

Industry studies find that Rust Belt producers were slow adopters (compared to producers elsewhere)

I

Two new technologies in steel of 1950s and 1960s: I

Basic oxygen furnace

I

Continuous casting methods

0

10

Percent 20 30

40

50

Fraction of Steel Made Using Continuous Casting Process

1968

1970

1972

1974

1976

1978

year US

Japan

Germany

Italy

France

Additional Evidence

Post-1980 Growth

Conclusion

Signed Confession

From 1980 Annual Report of American Iron and Steel Institute: “Inadequate capital formation in any industry produces meager gains in productivity, upward pressure on prices, sluggish job creation, and faltering economic growth. These effects have been magnified in the steel industry. Inadequate capital formation ... has prevented adequate replacement and modernization of steelmaking facilities, thus hobbling the industry’s productivity and efficiency.”

Did Productivity Growth Pick up After 1980s? I

I

Steel: I

US vertically integrated mills (mostly in Rust Belt)

I

11 percent TFP growth from 1982 to 1987; 16 percent 1992 to 1997

I

Source: Collard-Wexler and De Loecker (2012), Lieberman and Johnson (1999)

Autos: I

Pick up seen in cars per worker at GM, Ford and Chrysler

I

From annual reports; most operations in Rust Belt

I

Working on TFP numbers Other Predictions

Conclusion