Download here

Productivity or Employment:
Is it a choice?
Andrea De Michelis
Federal Reserve Board
Marcello Estevão
International Monetary Fund
Beth Anne Wilson
Federal 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.
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.
2
Take the case of Canada:
TFP growth has been particularly low.
G-7 Countries*: Growth in Total Factor Productivity
(percent)
4
4
Range of TFP Growth
Observed in the G-7
3
3
Average G-7
2
2
1
1
0
0
Canada
-1
1960
-1
1970
1980
1990
2000
* 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 to
1987 is for West Germany. Source: Organisation for Economic Co-Operation and Development
3
In contrast, employment growth
has been quite strong.
G-7 Countries*: Employment Growth
(percent)
9
9
6
6
Canada
3
3
0
0
-3
-3
Range of employment
growth rates observed
in the G-7
Average G-7
-6
-6
1950
1960
1970
1980
1990
2000
* 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.
4
Same with hours of work.
G-7 Countries*: Total Hours Worked
(percent change)
10
10
Average G-7
Canada
5
5
0
0
-5
-5
Range of Growth
in Hours Worked
Observed in the G-7
-10
-10
1950
1960
1970
1980
1990
2000
* 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.
5
As a result, Canadian GDP growth has
outperformed the G7 average.
G-7 Countries*: Real GDP** Growth
(percent)
14
14
Range of growth
rates observed in the G-7
Canada
7
7
0
0
Average G-7
-7
-7
1950
1960
1970
1980
1990
2000
* 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.
6
More generally, growth in TFP and labor input
are negatively correlated across the OECD.
TFP Growth and Hours Growth
TFP Growth (percent; average over 1970-2007)
2
Finland
1.5
Germany
Austria
Norway
TFP = -0.49H + 1.07
R² = 0.48
United Kingdom
France
Italy
1
Belgium
Netherlands
Sweden
Spain
United States
Japan
Switzerland
Australia
0.5
Denmark
Greece
Portugal
Canada
0
-1
-0.5
0
0.5
1
1.5
2
New Zealand
-0.5
Hours Growth (percent; average over 1970-2007)
Source: Total Economy Database.
7
Data: Labor Input and TFP



The Conference Board Total Economy
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)
8
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 Database

9
Negative correlation of TFP and hours
growth is robust, holding across datasets
and labor inputs...
KLEMS†
Database
TED
Labor Input Employment Employment
Constant
Coefficient
1.35***
(0.17)
-0.53***
(0.15)
0.86***
(0.18)
-0.36*
(0.17)
TED
Hours
KLEMS†
Hours
1.07***
(0.10)
-0.49***
(0.11)
0.74***
-0.12
-0.37**
(-0.09)
Observations
20
14
20
14
Adjusted 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
10
…and across time.
Correlation remains negative and significant
decade by decade (except 90s.)
Hours Growth vs. TFP Growth
Period
Constant
Hours
Growth
Observations
Adjusted R 𝟐
1970-2007
1970s
1980s
1990s
2000-2007
1.07***
(0.10)
1.67***
(0.13)
1.01***
(0.13)
0.60***
(0.15)
0.91***
(0.22)
-0.49***
(0.11)
-0.57***
(0.13)
-0.41***
(0.13)
-0.19
(0.18)
-0.63***
(0.18)
20
0.48
20
0.49
20
0.33
20
0.01
20
0.36
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
11
Countries relative relationship between
TFP and H growth fairly stable.
1970-2007
2.5
Above average TFP
Below average Hours
Above average TFP
Above average Hours
2.0
FIN
DEU
AUT
GBR
NLD
FRA
ITA
ESP
BEL
JPN
1.5
NOR
Average TFP Growth = 0.84
USA
SWZ SWE
DNK
1.0
0.5
AUS
PRT
GRC
CAN
0.0
NZL
-0.5
Average Hours Growth = 0.47
Below average TFP
Above average Hours
Below average TFP
Below average Hours
-1.0
-1.5
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
2.5
Dotted lines represent averages over 1970-2007.
12
But, some drift toward lower TFP/
stronger hours growth in Europe.
(1970s and 1980s)
1970s
1980s
AUT
ITA
BEL NLD ESP
FRA FIN NOR
DEU
2.5
2.5
2.0
2.0
GRC
ESP
1.5
JPN
AUS
SWZ
1.0
USA
SWE
CAN
0.5
0.0
NLZ
GBR FIN
FRA
TFP Growth (percent)
TFP Growth (percent)
3.0
PRT
GBR
DNK
3.0
BEL
ITA
DEU
1.5
NOR
PRT
JPN
NLD
AUT
GRC
1.0
USA
0.5
SWE
SWZ CAN
DNK
AUS
0.0
NLZ
-0.5
-0.5
-1.0
-1.0
-1.5
-1.5
-2.0
-2.0
-2.0
-1.0
0.0
1.0
Hours Growth (percent)
2.0
3.0
-2.0
Dotted lines represent the averages over 1970-2007 on all charts.
-1.0
0.0
1.0
Hours Growth (percent)
2.0
3.0
13
1990s and 2000-2007
1990s
2000-2007
NOR
2.5
2.5
2.0
2.0
FIN
1.5
AUT
DEU
SWE
GBR
ITA
FRA
JPN
3.0
AUS
1.0
NLD
DNK
PRT
BEL
SWZ CAN
GRC
USA
0.5
NLZ
0.0
ESP
DEU
TFP Growth (percent)
TFP Growth (percent)
FIN
3.0
1.5
SWE
AUT
GBR SWZ
JPN
NLD USA
FRA
GRC
BEL
DNK
CAN
1.0
0.5
NLZ
0.0
NOR
ITA
-0.5
-0.5
AUS
-1.0
-1.0
-1.5
-1.5
PRT
-2.0
-2.0
-1.0
0.0
1.0
Hours Growth (percent)
2.0
3.0
-2.0
-2.0
-1.0
0.0
1.0
Hours Growth (percent)
2.0
3.0
Dotted lines represent the averages over 1970-2007 on all charts.
14
Correlation of growth in TFP and hours
varies by sector (OECD 14)
2
Industry
Coefficient
Constant
Hotels and Restaurants
Manufacturing
-0.60** (0.26)
-0.46 (0.35)
0.28 (0.49)
1.19** (0.49)
14
14
0.25
0.05
-0.37**
-0.35*
-0.33
-0.23*
-0.23
(0.14)
(0.19)
(0.48)
(0.12)
(0.26)
0.74***
0.11
1.31***
0.39
0.81**
(0.12)
(0.30)
(0.40)
(0.41)
(0.30)
14
14
14
14
14
0.33
0.15
-0.04
0.18
-0.02
-0.21
-0.15
-0.13
-0.11
(0.31)
(0.19)
(0.28)
(0.37)
2.77***
0.24
0.43
1.37***
(0.81)
(0.25)
(1.04)
(0.42)
14
14
14
14
-0.04
-0.03
-0.06
-0.08
Total Economy
Other Services
Wholesale and Retail
Financial Services
Electricity
Agriculture, Forestry, and
Fishing
Construction
Mining and Quarrying
Transportation
Observations Adjusted R
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
Source: World KLEMS, EU KLEMS.
15
But variation in industry composition does
not explain cross-country variance.
TFP Growth vs. Hours Growth
U.S. time-varying
Baseline
weight
Constant
Hours Growth
0.74***
(0.12)
-0.37**
(0.14)
0.82***
(0.13)
-0.38***
(0.09)
Observations
14
14
Adjusted R2
0.33
0.56
Standard errors in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
Sources: EU KLEMS, World KLEMS and authors’ calculations.
16
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”.
17
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?
18
Causality: strategy
• 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.
19
Causality: IV regressions using tax wedge and
population growth
Step 1- Dependent variable -- Hours Growth
Constant
-2.42**
(1.00)
Average Tax Wedge
4.52**
(1.60)
Population Growth
Step 2- Dependent variable -- TFP Growth
Constant
1.22***
(0.11)
Predicted Hours
Growth
-0.71***
(0.19)
Observations
15
-0.55***
(0.16)
1.80***
(0.24)
1.07***
(0.11)
-0.47***
(0.15)
20
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
Time period spans 1970-2007.
Sources: Total Economy Database, McDaniel tax data.
20
Is the instrument independently correlated
with TFP growth?
Dependent variable -- TFP Growth
Constant
Hours Growth
Average Tax Wedge
2.18***
(0.54)
-0.31**
(0.12)
-1.79*
(0.90)
Population Growth
Observations
1.03***
(0.21)
-0.53**
(0.24)
0.10
(0.49)
15
20
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
Time period spans 1970-2007.
Sources: Total Economy Database, McDaniel tax data.
21
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.
22
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.
23
Thanks!
24
G-7 Countries*: Real GDP** Growth
(percent)
14
14
Canada
France
Germany
Italy
United Kingdom
United States
Japan
7
7
0
0
-7
-7
1950
1960
1970
1980
1990
2000
* 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.
25
TFP Growth vs. Hours Growth
by Sector (G7)
Industry
Coefficient
Constant
Observations
Adjusted R2
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
1.12*** (0.19)
7
0.73
Manufacturing
-0.48*** (0.12)
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.
26
Results: Using Tax Wedge as an Instrument
for Hours
Step 1 Regression
Hours Growth vs. Average Tax Wedge† by Period
Decades
1970-2007 1970s
1980s
1990s
Constant
Average Tax Wedge
-2.42**
(1.00)
4.52**
(1.60)
-4.21**
(1.82)
6.15**
(2.69)
-2.88*
(1.35)
5.51**
(2.14)
-1.61
(1.19)
3.23
(1.96)
2000-2007
-0.23
(1.33)
1.82
(2.21)
Observations
15
15
15
15
15
Adjusted 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)
2
Step 2 Regression
TFP Growth vs. Predicted Hours Growth by Period
Decades
1970-2007 1970s
1980s
1990s
Constant
Predicted Hours Growth
Observations
Adjusted R 2
2000-2007
1.22***
(0.11)
-0.71***
(0.19)
1.73***
(0.16)
-0.83***
(0.27)
1.08***
(0.20)
-0.37
(0.26)
0.75***
(0.17)
-0.73*
(0.37)
1.46
(0.84)
-1.13
(0.97)
15
0.49
15
0.37
15
0.07
15
0.17
15
0.09
27
TFP Growth vs. Hours Growth and Average Tax Wedge
Periods
1970-2007 1970s
Constant
Hours Growth
1990s
2000-2007
2.18***
3.55***
0.63
1.69*
1.59*
(0.54)
(1.09)
(0.67)
(0.79)
(0.79)
-0.14
-0.56***
-0.31**
Average Tax Wedge
1980s
-0.40** -0.53***
(0.12)
(0.14)
(0.12)
(0.17)
(0.17)
-1.79*
-2.66
0.86
-1.89
-1.03
(0.90)
(1.60)
(1.11)
(1.34)
(1.35)
15
15
15
15
15
0.64
0.60
0.63
0.14
0.46
2
Observations
Adjusted R
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
28
Results: Using Population Growth as an
Instrument for Hours
Step 1 Regression
Hours Growth vs. Population Growth by Decade
Decades
1970-2007 1970s 1980s 1990s
Constant
Population Growth
Observations
2
Adjusted R
-0.55*** -1.31*** -0.15
-0.27
(0.16)
(0.28) (0.23) (0.31)
1.80*** 1.96*** 1.58*** 1.22**
(0.24)
(0.36) (0.38) (0.46)
20
0.75
20
0.61
20
0.46
20
0.24
2000-2007
0.12
(0.17)
1.58***
(0.27)
20
0.64
Step 2 Regression
TFP Growth vs. Predicted Hours Growth by Decade
Decades
1970-2007 1970s 1980s 1990s 2000-2007
Constant
Predicted Hours Growth
Observations
Adjusted R 2
1.07*** 1.67*** 0.97*** 0.53**
(0.11)
(0.16) (0.18) (0.20)
-0.47*** -0.52** -0.34
-0.02
(0.15)
(0.20) (0.21) (0.34)
20
0.33
20
0.24
20
0.07
20
-0.06
0.90***
(0.28)
-0.62**
(0.25)
20
0.21
29
TFP Growth vs. Hours Growth and Population Growth
Decade
1970-2007 1970s
Constant
1980s
1.03*** 1.52*** 0.94***
(0.21)
Hours Growth
(0.39)
(0.18)
-0.53** -0.63*** -0.48**
Population Growth
1990s 2000-2007
0.47
0.90***
(0.28)
(0.24)
-0.25
-0.65*
(0.24)
(0.22)
(0.18)
(0.21)
(0.33)
0.10
0.22
0.22
0.28
0.04
(0.49)
(0.54)
(0.41)
(0.48)
(0.63)
20
20
20
20
20
0.45
0.46
0.31
-0.03
0.32
2
Observations
Adjusted R
Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1
30