Productivity or Employment: Is it a choice?

Andrea De Michelis Federal Reserve Board

Marcello EstevãoInternational Monetary Fund

Beth Anne WilsonFederal Reserve Board

January 4, 2013

The views in this presentation are solely the responsibility of the

authors and should not be interpreted as reflecting the views of the International Monetary Fund or

the Board of Governors of the Federal Reserve System or of any other

person associated with the Federal Reserve System. 

3

Background

In general, economic theory assumes that TFP growth follows an exogenous process.

Low TFP growth is seen as worrisome, as many associate it with poor economic performance.

In reality, not a one-to-one relation between TFP and output growth key motivation for this paper: TFP growth may be a “choice” variable.

4

Take the case of Canada: TFP growth has been

particularly low.

-1

0

1

2

3

4

-1

0

1

2

3

4

1960 1970 1980 1990 2000

G-7 Countries*: Growth in Total Factor Productivity(percent)

Canada

Average G-7

Range of TFP Growth Observed in the G-7

* Data start in 1961 for the United States, 1962 for Germany and Italy, 1963 for Canada, 1970 for Japan, and 1971 for France and United Kingdom. German data prior to1987 is for West Germany. Source: Organisation for Economic Co-Operation and Development

5

In contrast, employment growth

has been quite strong.

-6

-3

0

3

6

9

-6

-3

0

3

6

9

1950 1960 1970 1980 1990 2000

G-7 Countries*: Employment Growth(percent)

Range of employmentgrowth rates observed

in the G-7

Canada

Average G-7

* Data for Canada, Italy, Japan, and United Kingdom start in 1960. German data prior to 1987 are for West Germany.Source: The Conference Board Total Economy Database, January 2010.

6

Same with hours of work.

-10

-5

0

5

10

-10

-5

0

5

10

1950 1960 1970 1980 1990 2000

G-7 Countries*: Total Hours Worked(percent change)

CanadaAverage G-7

Range of Growthin Hours Worked

Observed in the G-7

* Data for Canada, Italy, Japan, and the United Kingdom start in 1960. Data for Germany prior to 1987 are for West Germany.Source: The Conference Board Total Economy Database, January 2010.

7

As a result, Canadian GDP growth has outperformed the G7 average.

-7

0

7

14

-7

0

7

14

1950 1960 1970 1980 1990 2000

G-7 Countries*: Real GDP** Growth(percent)

Canada

Average G-7

Range of growthrates observed in the G-7

* German data prior to 1987 are for West Germany. ** Converted to constant 1990 dollars using Geary-Khamis PPP weights.Source: The Conference Board Total Economy Database, January 2010.

8

More generally, growth in TFP and labor input are negatively

correlated across the OECD.

-1 -0.5 0 0.5 1 1.5 2

-0.5

0

0.5

1

1.5

2

Australia

Austria

Belgium

Canada

Denmark

Finland

France

Germany

Greece

Italy

Japan

Netherlands

New Zealand

Norway

Portugal

SpainSweden

Switzerland

United Kingdom

United States

f(x) = − 0.486014503646795 x + 1.07207926602639R² = 0.503239472566177

TFP Growth and Hours Growth

Hours Growth (percent; average over 1970-2007)

TF

P G

row

th (

perc

ent;

aver

age

over

197

0-20

07)

Source: Total Economy Database.

9

Data: Labor Input and TFP The Conference Board Total Econom

y Database: total economy annual data, main 20 OECD countries, 1970-2007

World/EU KLEMS: annual data, 14 OECD countries, 10 sectors, various sample ranges, but 1980-2007 available for most countries of interest

EU AMECO: total economy annual data, European and other G-7 countries, 1960-2013(no hours data, used only for robustness analysis)

10

Other Data

Sources for tax data McDaniel (2007): payroll, income, and

consumption taxes, 15 OECD countries, 1950/70-2007

Sources for population data United Nations The Conference Board Total Economy D

atabase

11

Negative correlation of TFP and hours growth is robust, holding across datasets and

labor inputs...Database TED KLEMS† TED KLEMS†Labor Input Employment Employment Hours Hours

Constant 1.35*** 0.86*** 1.07*** 0.74***(0.17) (0.18) (0.10) -0.12

Coefficient -0.53*** -0.36* -0.49*** -0.37**(0.15) (0.17) (0.11) (-0.09)

Observations 20 14 20 14Adjusted R2 0.36 0.21 0.48 0.33†KLEMS data spans the time period 1980-2007.Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

12

…and across time. Correlation remains negative and

significant decade by decade (except 90s.)

Hours Growth vs. TFP GrowthPeriod 1970-2007 1970s 1980s 1990s 2000-2007

           

Constant 1.07*** 1.67*** 1.01*** 0.60*** 0.91***

(0.10) (0.13) (0.13) (0.15) (0.22)

Hours Growth -0.49*** -0.57*** -0.41*** -0.19 -0.63***

(0.11) (0.13) (0.13) (0.18) (0.18)

Observations 20 20 20 20 20

Adjusted R 0.48 0.49 0.33 0.01 0.36

Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

𝟐

13

-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

AUS

AUT

BEL

CAN

DNK

FIN

FRA

DEU

GRC

ITA

JPN

NLD

NZL

NOR

PRT

ESP

SWESWZ

GBR

USA

1970-2007

Above average TFPBelow average Hours

Below average TFPBelow average Hours

Above average TFPAbove average Hours

Below average TFPAbove average Hours

Average TFP Growth = 0.84

Average Hours Growth = 0.47

Dotted lines represent averages over 1970-2007.

Countries relative relationship between TFP and H growth

fairly stable.

14

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

AUS

AUT

BEL

CAN

DNK

FINFRADEU

GRC

ITA

JPN

NLD

NLZ

NOR

PRT

ESP

SWE

SWZ

GBR

USA

1970s

Hours Growth (percent)

TFP

Gro

wth

(per

cent

)But, some drift toward lower

TFP/ stronger hours growth in Europe.

(1970s and 1980s)

Dotted lines represent the averages over 1970-2007 on all charts.

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

AUS

AUT

BEL

CAN

DNK

FINFRA

DEU

GRC

ITA

JPNNLD

NLZ

NOR

PRT

ESP

SWE

SWZ

GBR

USA

1980s

Hours Growth (percent)

TFP

Gro

wth

(per

cent

)

15

1990s and 2000-2007

Dotted lines represent the averages over 1970-2007 on all charts.

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

AUS

AUT

BEL

CAN

DNK

FIN

FRA

DEU

GRC

ITA

JPN

NLD

NLZ

NOR

PRT

ESP

SWE

SWZ

GBR USA

1990s

Hours Growth (percent)

TFP

Gro

wth

(per

cent

)

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

AUS

AUT

BELCANDNK

FIN

FRA

DEU

GRC

ITA

JPN NLD

NLZ

NOR

PRT

SWE

SWZGBR

USA

2000-2007

Hours Growth (percent)

TFP

Gro

wth

(per

cent

)

16

Correlation of growth in TFP and hours varies by sector

(OECD 14)Industry Coefficient Constant Observations Adjusted R

Hotels and Restaurants -0.60** (0.26) 0.28 (0.49) 14 0.25

Manufacturing -0.46 (0.35) 1.19** (0.49) 14 0.05

Total Economy -0.37** (0.14) 0.74*** (0.12) 14 0.33

Other Services -0.35* (0.19) 0.11 (0.30) 14 0.15

Wholesale and Retail -0.33 (0.48) 1.31*** (0.40) 14 -0.04

Financial Services -0.23* (0.12) 0.39 (0.41) 14 0.18

Electricity -0.23 (0.26) 0.81** (0.30) 14 -0.02Agriculture, Forestry, and

Fishing -0.21 (0.31) 2.77*** (0.81) 14 -0.04

Construction -0.15 (0.19) 0.24 (0.25) 14 -0.03

Mining and Quarrying -0.13 (0.28) 0.43 (1.04) 14 -0.06

Transportation -0.11 (0.37) 1.37*** (0.42) 14 -0.08                     

Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

Source: World KLEMS, EU KLEMS.

2

17

But variation in industry composition does not explain

cross-country variance.

2

TFP Growth vs. Hours Growth

BaselineU.S. time-varying

weight

Constant 0.74*** 0.82***(0.12) (0.13)

Hours Growth -0.37** -0.38***(0.14) (0.09)

Observations 14 14Adjusted R   0.33   0.56  

Standard errors in parentheses.*** p<0.01, ** p<0.05, * p<0.1Sources: EU KLEMS, World KLEMS and authors’ calculations.

18

What could explain this negative correlation?

• Measurement error? Probably not. Measurement issues with TFP more relevant at cyclical frequencies.

Result does not depend on the database used (TED, World/EU KLEMS)

Country mix of TFP and hours growth is relatively stable over time.

Result holds within industry/country pair.

• TFP as a choice variable: Given the availability of labor inputs, TFP growth is “chosen”.

19

Causality: hypothesis • Factor endowment not only affects the choice of capital or labor-intensive technologies but also how much to invest in techniques and processes that boost TFP.

• Given that productivity innovations are costly, countries with abundant labor supply may “choose” less productivity growth.

• Test: Is there a causality going from labor supply shocks to TFP growth?

Causality: strategy

20

• Find variables that affect TFP growth only through the decision of hiring labor.• Use these variables as instruments in regressions linking TFP growth to hours growth.• Good candidates:

Tax wedge: differences in taxes influence labor supply and introduce a gap between MRS and MPL (Prescott, 2004, and Ohanian et al., 2007).

Population growth: availability of labor input.

21

Causality: IV regressions using tax wedge and population growth

Step 1- Dependent variable -- Hours GrowthConstant -2.42** -0.55***

(1.00) (0.16)Average Tax Wedge 4.52**

(1.60)Population Growth 1.80***

(0.24)Step 2- Dependent variable -- TFP Growth

Constant 1.22*** 1.07***(0.11) (0.11)

Predicted Hours Growth -0.71*** -0.47***

(0.19) (0.15)Observations 15 20Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

Time period spans 1970-2007.Sources: Total Economy Database, McDaniel tax data.

22

Is the instrument independently correlated with TFP growth?

Dependent variable -- TFP Growth

Constant 2.18*** 1.03***(0.54) (0.21)

Hours Growth -0.31** -0.53**(0.12) (0.24)

Average Tax Wedge -1.79*(0.90)

Population Growth 0.10(0.49)

Observations 15 20Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

Time period spans 1970-2007.

Sources: Total Economy Database, McDaniel tax data.

23

Conclusions

There is robust negative correlation between TFP growth and hours growth across OECD countries.

At least some of this negative correlation seems to be a result of reactions to shocks in labor input. So, TFP could, in part, be a “choice” variable.

This mechanism makes more sense than explanations of TFP growth differences between, say, Germany and Canada based on institutions. These are all rich, mature societies with good institutions.

The endogeneity of TFP could also help explain longer-run developments in Europe and Canada.

24

Conclusions

Looking ahead, population aging could trigger a wage adjustment and an endogenous increase in TFP growth in countries so far specialized in fast hours growth/low TFP growth. But no guarantee, look at Japan.

Good institutions that support innovation and product market competition are always good for TFP growth, and would raise incentives to be more productive and ease transition.

25

Thanks!

26

-7

0

7

14

-7

0

7

14

1950 1960 1970 1980 1990 2000

G-7 Countries*: Real GDP** Growth(percent)

Canada France

Germany Italy

United Kingdom United States

Japan

* German data prior to 1987 are for West Germany. ** Converted to constant 1990 dollars using Geary-Khamis PPP weights.Source: The Conference Board Total Economy Database, January 2010.

27

TFP Growth vs. Hours Growth by Sector (G7)

Industry Coefficient Constant Observations Adjusted R

Hotels and Restaurants -0.99** (0.27) 1.07* (0.50) 7 0.67

Other Services -0.72 (0.36) 0.72 (0.50) 7 0.33

Manufacturing -0.48*** (0.12) 1.12*** (0.19) 7 0.73

Wholesale and Retail -0.49 (0.48) 1.74*** (0.39) 7 0.01

Total Economy -0.47** (0.15) 0.78*** (0.11) 7 0.59

Electricity -0.42 (0.44) 0.41 (0.47) 7 -0.02

Construction -0.35 (0.38) 0.02 (0.43) 7 -0.03

Mining and Quarrying -0.18 (0.24) -1.20 (0.96) 7 -0.08

Agriculture, Forestry, Fishing -0.17 (0.66) 3.06 (1.86) 7 -0.19

Transportation 0.16 (0.69) 0.98 (0.79) 7 -0.19

Financial Services -0.16 (0.19) 0.075 (0.60) 7 -0.06                     

Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

Source: World KLEMS, EU KLEMS.

2

28

Results: Using Tax Wedge as an Instrument for HoursStep 1 Regression          

Hours Growth vs. Average Tax Wedge† by PeriodDecades 1970-2007 1970s 1980s 1990s 2000-2007           Constant -2.42** -4.21** -2.88* -1.61 -0.23  (1.00) (1.82) (1.35) (1.19) (1.33)Average Tax Wedge 4.52** 6.15** 5.51** 3.23 1.82  (1.60) (2.69) (2.14) (1.96) (2.21)           Observations 15 15 15 15 15Adjusted R 0.33 0.23 0.29 0.11 -0.02

† Equal to (1- tax rate on labor income)/(1 + tax rate on consumption expenditures)

           

Step 2 Regression          

TFP Growth vs. Predicted Hours Growth by PeriodDecades 1970-2007 1970s 1980s 1990s 2000-2007           Constant 1.22*** 1.73*** 1.08*** 0.75*** 1.46  (0.11) (0.16) (0.20) (0.17) (0.84)Predicted Hours Growth -0.71*** -0.83*** -0.37 -0.73* -1.13  (0.19) (0.27) (0.26) (0.37) (0.97)           Observations 15 15 15 15 15Adjusted R 0.49 0.37 0.07 0.17 0.09

2

2

29

TFP Growth vs. Hours Growth and Average Tax WedgePeriods 1970-2007 1970s 1980s 1990s 2000-2007

           

Constant 2.18*** 3.55*** 0.63 1.69* 1.59*

  (0.54) (1.09) (0.67) (0.79) (0.79)

Hours Growth -0.31** -0.40** -0.53*** -0.14 -0.56***

  (0.12) (0.14) (0.12) (0.17) (0.17)

Average Tax Wedge -1.79* -2.66 0.86 -1.89 -1.03

  (0.90) (1.60) (1.11) (1.34) (1.35)

           Observations

15 15 15 15 15

Adjusted R 0.64 0.60 0.63 0.14 0.46

Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1  

2

30

Step 1 RegressionHours Growth vs. Population Growth by Decade

Decades 1970-2007 1970s 1980s 1990s 2000-2007           

Constant -0.55*** -1.31*** -0.15 -0.27 0.12(0.16) (0.28) (0.23) (0.31) (0.17)

Population Growth 1.80*** 1.96*** 1.58*** 1.22** 1.58***(0.24) (0.36) (0.38) (0.46) (0.27)

Observations 20 20 20 20 20Adjusted R 0.75 0.61 0.46 0.24 0.64

Step 2 RegressionTFP Growth vs. Predicted Hours Growth by Decade

Decades 1970-2007 1970s 1980s 1990s 2000-2007           Constant 1.07*** 1.67*** 0.97*** 0.53** 0.90***

(0.11) (0.16) (0.18) (0.20) (0.28)Predicted Hours Growth -0.47*** -0.52** -0.34 -0.02 -0.62**

(0.15) (0.20) (0.21) (0.34) (0.25)

Observations 20 20 20 20 20Adjusted R 0.33 0.24 0.07 -0.06 0.21

2

2

Results: Using Population Growth as an Instrument for Hours

31

TFP Growth vs. Hours Growth and Population Growth

Decade 1970-2007 1970s 1980s 1990s 2000-2007

           

Constant 1.03*** 1.52*** 0.94*** 0.47 0.90***

  (0.21) (0.39) (0.18) (0.28) (0.24)

Hours Growth -0.53** -0.63*** -0.48** -0.25 -0.65*

  (0.24) (0.22) (0.18) (0.21) (0.33)

Population Growth 0.10 0.22 0.22 0.28 0.04

  (0.49) (0.54) (0.41) (0.48) (0.63)

           Observations

20 20 20 20 20

Adjusted R 0.45 0.46 0.31 -0.03 0.32

Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1  

2