DECOUPLING OF WAGES FROM PRODUCTIVITY

Chiara Criscuolo & Cyrille Schwellnus

IAB Workshop, 17 December 2018

2

Productivity gains no longer translate into broadly shared wage gains

Note: Employment weighted average of 24 countries (two-year moving averages ending in the indicated years). 1995-2013 for Finland,

Germany, Japan, Korea, United States; 1995-2012 for France, Italy, Sweden; 1996-2013 for Austria, Belgium; United Kingdom; 1996-2012 for

Australia, Spain; 1997-2013 for Czech Republic, Denmark, Hungary; 1997-2012 for Poland; 1998-2010 for Netherlands; 1998-2013 for

Norway; 1998-2012 for Canada, New Zealand; 1999-2013 for Ireland; 2002-2011 for Israel; 2003-2013 for Slovak Republic. All series are

deflated by the value added price index excluding the primary, housing and non-market sectors.

Source: OECD Economic Outlook November 2018.

100

105

110

115

120

125

130

135

1995 1997 1999 2001 2003 2005 2007 2009 2011 2013

Index 1995 = 100

Labour productivity Average wages Median wages

Contribution of declining labour share

Contribution of increased wage inequality

3

Large heterogeneity in decoupling across countries, 1995-2013

Note: Annual growth rates in %; excluding primary, housing and non-market industries

Source: OECD Economic Outlook November 2018

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0

Belgium

Japan

Slovak Republic

Canada

Netherlands

Australia

Israel

Ireland

United States

Hungary

Korea

Poland

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0

Italy

Finland

Spain

France

United Kingdom

Denmark

Sweden

Czech Republic

New Zealand

Germany

Austria

Norway

Total decoupling Labour share Wage inequality

Annualised growth rates in %; excluding primary, housing and non-market industries; 1995-2014

4

“Superstar” firms or the rest?

Labour productivity and real wages (2001 = 100)

Panel A: Countries with declines in labour shares Panel B: Countries with increases in labour shares

90

100

110

120

130

140

150

160

2001 2003 2005 2007 2009 2011 2013 2001 2003 2005 2007 2009 2011 2013

Note: Labour productivity and real wages are computed as the unweighted mean across firms of real value added per worker and real labour compensation per worker.

Leaders are defined as the top 5% of firms in terms of labour productivity within each country group in each industry and year. The countries with a decline in the labour

share excluding the primary, housing, financial and non-market industries over the period 2001-2013 are: Belgium, Denmark, Germany, Ireland, Japan, Korea,

Sweden, United Kingdom and United States. The countries with an increase are: Austria, Czech Republic, Estonia, Finland, France, Italy, Netherlands and Spain.

Source: OECD Economic Outlook November 2018.

(including Germany)

5

Between-firm wage dispersion has increased

90-10 percentile ratio (2001=100)

Note: The solid and dashed lines are based on the estimated year dummies of a regression of, respectively, log-productivity and wage dispersion across firms within country-

sector pairs in the following countries: Australia, Austria, Belgium, Chile, Denmark, Finland, France, Hungary, Italy, Japan, Netherlands, New Zealand, Norway, and Sweden.

The dotted line is based on the year dummy estimates of a regression of the worker-level wage dispersion from the OECD Earnings Distribution database within each country

(Australia, Finland, France, Hungary, Italy, Japan, Netherlands, New Zealand, Norway, and Sweden).

Source: Berlingieri, Blanchenay and Criscuolo (2017) and authors’ calculations.

98

100

102

104

106

108

110

112

114

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Worker-level wage dispersion Between-firm wage dispersion Between-firm labour productivity dispersion

6

Labour share analysis: Descriptive stats

Note: GDP weighted average of 20 OECD countries included in the industry-level regressions. The black lines indicate cumulated changes;

the red lines indicate the corresponding trends; and the dotted lines indicate +/- 1 standard deviation around the weighted average.

Source: OECD National Accounts Database and OECD TiVA Database.

-9

-6

-3

0

3

1995 2000 2005 2010

A. Labour share

% points

-30

-20

-10

0

10

1995 2000 2005 2010

B. Relative investment price

%

-10

-5

0

5

10

15

20

1995 2000 2005 2010

C. GVC participation

% points

7

Labour share analysis: Descriptive stats

Changes in public policies and institutions1, 1995-2011

-4

-3

-2

-1

0

1

2

3

4

PMR Corporatetax

ALMP CBcoverage

Taxwedge

EPL Minimumwage

Centralisationof CB

25th to 75th percentiles Minimum Median Maximum

USA

EST

ESP

FIN

KOR

AUS KOR

AUT

CZE

DEU SWE

DEU IRL KOR IRL

FRA

1 Normalised by the cross-country standard deviation in the initial year

8

Labour share analysis: Descriptive stats

Changes in labour shares overwhelmingly reflect within-industry developments

Note: Based on a shift-share decomposition of aggregate changes in labour shares.

Source: OECD National Accounts Database.

AUT

BEL

CZEDNK

EST

FINFRA

DEU

IRL

ITA

KOR

NLD

NOR

SVK

ESPSWE

GBR

USA

-15

-10

-5

0

5

10

-15 -10 -5 0 5 10

Cha

nge

in a

ggre

gate

labo

ur s

hare

(in

% p

oint

s)

Contribution of within industry change (in % points)

45

9

Labour share analysis: Results

∆𝐿𝑆𝑖𝑗𝑡 = 𝛽1∆𝑃𝑖𝑗𝑡𝐼𝑛𝑣 + 𝛽2∆𝑇𝑖𝑗𝑡 + 𝛽3 𝑅𝑇𝐼𝑖𝑗𝑡

0 × ∆𝑃𝑖𝑗𝑡𝐼𝑛𝑣 +𝛽4 𝐸𝑥𝑝𝑗

𝑘 × ∆𝑃𝑜𝑙𝑖𝑡𝑘 + 𝛽5𝑋𝑖𝑗𝑡 + 𝛼𝑖 + 𝛼𝑗𝑡 + 휀𝑖𝑗𝑡

10

High skills reduce capital-labour substitution even in high-routine industries

Change in the labour share in response to a 10% decrease in the

relative investment price, % points

-4

-3

-2

-1

0

1

Low routineHigh skill

High routineHigh skill

Low routineLow skill

High routineLow skill

90% confidence interval90% confidence interval

11

Labour share analysis: Results

Ratio of average

wages to labour

productivity

Ratio of median to average

wages or ratio of bottom to

top firm-level wages

Labour share1

Inverse measure of wage

inequality2

Technological change mTrade integration mHigh skills kCompetition-friendly product market reform kLoosening of employment protection kMinimum wage reduction kCollective bargaining decentralisation XALMP spending increase k

12

Between-firm wage dispersion has increased

90-10 percentile ratio (2001=100)

98

100

102

104

106

108

110

112

114

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Worker-level wage dispersion Between-firm wage dispersion Between-firm labour productivity dispersion

Between-firm wage dispersion

Worker-level

Between firm LP dispersion

Is increasing earnings inequality linked to growing

productivity dispersion among firms?

Several reasons why the wage bill would be higher in more

productive firms:

Rent-sharing from:

Asymmetric information

Wage bargaining

Sorting and assortative matching

structural factors, e.g. globalisation, technology/digitalisation

policies min. wage, EPL, union, coordinated wage setting

The task requires data representative for the entire

distribution of firms: MultiProd project across 16 countries

Mechanisms and drivers

Answer to come…new OECD project exploiting Matched employer-employee data

A distributed microdata approach

COUNTRIES

Policy questions Data strategy to provide evidence

in cross-country setting

Code for data cleaning, linking

and analysis

modular&

customised

CO

NFI

DEN

TIA

LM

ICR

OD

ATA

Metadata:variables, coverage,

weighting...

Code run by national experts

Agreed, non confidential microaggregated data

Descriptive and econometric evidence

Country notes

Measures of Productivity: LP; MFPR Wooldridge;

MFPR Solow…

Changes in distributions over time (productivity; wage

and size).

Firm-level productivity and employment growth ;

Entry and exit, job creation/destruction;

OP-Gap; Dynamic decomposition of OP-gap (Melitz

and Polanec, 2015);

Decomposition of aggregate MFP growth (Petrin and

Levinsohn, 2012);

Hsieh-Klenow Misallocation

….

Data: The OECD MultiProd project

Divergence more pronounced for the bottom half of the wage distribution

Wage dispersion comes mostly from the bottom of

the distribution both in manufacturing and services

Productivity Divergence more marked at the

bottom of the distribution

Econometric analysis shows that the two are strongly and significantly related

How much of the increase in Between-Firm Wage

dispersion is due to divergence in productivity?

Labour Productivity

MultiFactorProductivity

19

Wage inequality analysis: Results on structural factors

(1) (2) (3) (4)

Dependent variable

Log Imports 0.02

(0.07)

Log Exports 0.15*

(0.09)

Share of ICT in fixed assets 0.13**

(0.06)

Share of high-skilled -0.07

(0.05)

Year fixed effects YES YES YES YES

Country x industry fixed effects YES YES YES YES

Observations 1,779 1,779 1,917 2,190

Number of countries 12 12 8 11

Adjusted R² 0.92 0.92 0.97 0.97

Log Wage (90-10)

𝑊𝐷𝑐𝑗𝑡 = 𝛼 + 𝛾𝑋𝑐𝑗𝑡 + 𝛿𝑍𝑐𝑗𝑡 + 𝑦𝑡 + 𝑧𝑐𝑗 + 휀𝑐𝑗𝑡

20

Wage inequality analysis: Results on labour market institutions and regulation

𝑊𝐷𝑐𝑗𝑡 = 𝛼 + 𝛾𝑋𝑐𝑗𝑡 + 𝛿𝑍𝑐𝑗𝑡 + 𝑦𝑡 + 𝑧𝑐𝑗 + 휀𝑐𝑗𝑡

21

Summary

1. Based on Schwellnus et al. (2018) and Pak and Schwellnus (2018).

2. Based on De Serres and Schwellnus (2018) and Berlingieri, Blanchenay and Criscuolo (2017).

Note: indicates statistical insignificance and ? indicates that drivers have not been subject to robust empirical analysis in the context of the

studies reviewed in this chapter.

22

Summary

Some decoupling on average but significant cross-country heterogeneity.

Technology-driven declines in relative investment prices and increased

global value chain participation partly explain the decoupling of wage

growth from productivity growth.

Public policies and institutions that affect the scope for capital-labour

substitution as well as the size and the distribution of producer rents can help

explain large differences in decoupling across countries.

Labour share declines have been particularly pronounced at the

technological frontier and wage dispersion between firms has increased,

which may reflect technology- and globalisation-induced “winner-takes-

most” dynamics.

23

Where do we go from here?

Labour Share Wage Inequality

Within Firms Between Firms

Innovation /

Technology.

adoption

Worker

composition Market

power

Firm-level Composition

Between-firm

dispersion in

measured

productivity:

Link between wages and

productivity

24

Where do we go from here?

• Need harmonized matched employer-employee

• Challenging

• But it allows to “properly decomposing” overall wage inequality in

“within”vs “between”

look at the mechanisms:

• Sorting and worker composition

• Rent sharing

• Outsourcing

• ….

• More work using “firm level” information from microaggregated data on

labour shares

• Distribution of labour share along the whole productivity/size

distribution

• Quantify contribution to aggregate changes in labour share from

different types of firms

• K/L substitution vs rents (mark-ups)

• Reallocation vs within-changes

25

Summary:

OECD (2018), “Chapter 2: Decoupling of wages from productivity: What implications for public policies,” OECD Economic Outlook, November 2018.

Country-level evidence:

De Serres, A. and C. Schwellnus (2018), “A general equilibrium (LM and PM reforms) perspective to inequality”, in Astarita, C. and G. D’Adamo (eds.), Inequality and Structural Reforms: Methodological Concerns and Lessons fromPolicy. Workshop Proceedings, European Economy Discussion Papers No. 71, European Commission, Brussels.

Schwellnus, C., A. Kappeler and P. Pionnier (2017), “Decoupling of wages from productivity: Macro-level facts”, OECD Economics Department Working Papers, No. 1373, OECD Publishing, Paris.

Schwellnus, C., A. Kappeler and P. Pionnier (2017), “The Decoupling of Median Wages from Productivity in OECD Countries”, International Productivity Monitor, Vol. 32.

Industry- and firm-level evidence:

Berlingieri, G., P. Blanchenay and C. Criscuolo (2017), “The great divergence(s)”, OECD Science, Technology and Industry Policy Papers, No. 39, OECD Publishing, Paris.

Pak, M. and C. Schwellnus (2018), “Labour share developments over the past two decades: The role of public policies”, OECD Economics Department Working Papers, OECD Publishing, forthcoming.

Schwellnus, C., et al. (2018), “Labour share developments over the past two decades: The role of technological progress, globalisation and “winner-takes-most” dynamics”, OECD Economics Department Working Papers, No. 1503, OECD Publishing, Paris.

References

…which is also found in wages

Wage dispersion between more ICT-intensive vs less ICT-intensive sectors 90-10 difference in log Wages

Source: Berlingieri, G., P. Blanchenay and C. Criscuolo (2017) and MultiProd (2017).

Note: we define “ICT” sectors above the median ICT level and non-ICT those below. The figure plots the year

dummy estimates t of a regression of log-wage dispersion (measured as the difference between the 90th and 10th

percentiles of log-wages) within country-sector pairs, using data from the following countries: AUS, AUT, BEL, CHL,

DNK, FIN, FRA, HUN, ITA, JPN, NLD, NOR, NZL, SWE.

-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

ICT vs Non-ICT wage dispersion

Non-ICT intensive

ICT-intensive

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

Overall between-firms wage dispersion