Structural Change and
Productivity Growth
--in Latin America, Asia, and Turkey
Dani Rodrik
Merih Celasun Memorial Lecture
December 2010
Structuralism is back



Not in the sense of distrust of markets or underestimation of the
role of incentives
But greater appreciation of the role that “economic structure”
plays in facilitating and constraining economic development
As evidenced by work on:
 Inter-sectoral and inter-firm gaps in productivity

The export-diversification challenge
 Innovation as “self-discovery” rather than R&D
 Structural change as engine of development
Developing economies are not just radially-shrunk versions of
advanced economies


“dualism”
Labor productivity gaps: Turkey
Labor productivity in relation to average productivity, 2005
6
5
4
3
2
1
0
agr
min
man
pu
con
wrt
tsc
fire
cspsgs
Dualism within sectors
Source: McKinsey Global Institute (2003)
-.14
Inter-sectoral productivity gaps diminish
over the course of development
BWA
IDN
-.16
MYS
IND
BOL
BRA
TURTHA
VEN
PER
-.18
TWN
PHL
NLD
COL
CHL
-.2
MEX
DNK
HKG
SGP
KOR
ESP
JPN
ITA
SWE USA
UKM
-.22
CRI
ARG
FRA
7
8
9
lnrgdpch
10
11
Coefficient of variation of (log) sectoral labor productivity against percapita income
How does inter-sectoral structural change
contribute to overall productivity growth?
Pt    i ,t k pi ,t   pi ,t  i ,t
i n
i n
“within”
“structural change”
The Latin American paradox
Productivity decomposition in Latin America across different periods
(annual growth rates)
1950 - 1975
1975 - 1990
1990 - 2005
-0.01
-0.005
Sectoral productivity
growth
Structural change
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
Based on data from Carmen Pages, ed., The Age of Productivity, IDB, 2010.
0.04
0.045
Implications


Post-1990 growth constrained by adverse structural
change
 “growth reducing structural change”
Consequence of economic liberalization: positive within
effects, negative overall effects?
 Empirical work on productivity consequences of trade
liberalization within manufacturing
 What happens if displaced labor ends up in sectors
with even lower productivity

Informality, traditional services, etc.
A more detailed, comparative look: the data
Latin America
Asia
High-income
ARG
BOL
BRA
CHL
COL
CRI
MEX
PER
VEN
HKG
IDN
IND
KOR
MYS
PHL
SGP
THA
TWN
DNK
ESP
FRA
ITA
JPN
NLD
SWE
UKM
USA
1
2
3
4
5
6
7
8
9
Sector
AGR
MIN
MAN
PU
CON
WRT
TSC
FIRE
CSPSGS
Full name
Agriculture
Mining
Manufacturing
Public utilities
Construction
Wholesale & retail trade
Transport & communication
Finance & business services
Government & public services
+ TUR
TUR not
included in
this data set.
I used data
from TUIK to
include TUR
in the
analysis.
Marcel P. Timmer and Gaaitzen J. de Vries (2007), “A Cross-Country Database For Sectoral Employment And Productivity
In Asia And Latin America, 1950-2005,” Groningen Growth and Development Centre Research Memorandum GD-98,
Groningen: University of Groningen, August 2007.
Questions




Is this something due to post-1990 global conjuncture?
How does Asia compare?
Where does Turkey stand in comparison to Latin
America and Asia?
How can we explain these patterns?
Basic results
Decomposing productivity change, 1990-2005
LAC
ASIA
sectoral productivity growth
HI
structural change
-0.015
-0.01
region
-0.005
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
overall productivity growth
sectoral productivity growth
structural change
LAC
1.35%
2.24%
-0.88%
ASIA
3.33%
2.81%
0.52%
HI
1.46%
1.54%
-0.09%
0.04
Basic results (weighted data)
Decomposing productivity growth, 1990-2005 (weighted data)
LAC
ASIA
sectoral productivity growth
HI
structural change
-0.50%
region
0.00%
0.50%
1.00%
1.50%
2.00%
2.50%
3.00%
3.50%
overall productivity growth
sectoral productivity growth
structural change
LAC
1.08%
1.05%
0.03%
ASIA
3.28%
2.12%
1.16%
HI
1.81%
1.92%
-0.11%
Countries ranked by:
Contribution of sectoral:
country
KOR
PER
CHL
SGP
MYS
TWN
BOL
IND
VEN
ARG
HKG
TUR
IDN
THA
CRI
MEX
PHL
BRA
COL
ETH
region
ASIA
LAC
LAC
ASIA
ASIA
ASIA
LAC
ASIA
LAC
LAC
ASIA
TURKEY
ASIA
ASIA
LAC
LAC
ASIA
LAC
LAC
AFRICA
total
0.038972
0.034072
0.029252
0.037079
0.040816
0.039907
0.008808
0.042316
-0.003542
0.023534
0.03272
0.031586
0.027799
0.030511
0.0125
0.01067
0.009455
0.004444
0.001849
0.018733
within
0.052947
0.038536
0.038205
0.037853
0.035871
0.03448
0.033657
0.032409
0.032048
0.029429
0.020182
0.017353
0.017228
0.013835
0.008725
0.008339
0.00809
0.006957
0.00529
0.003935
Contribution of structural change:
structural
-0.013975
-0.004465
-0.008953
-0.000775
0.004946
0.005427
-0.024849
0.009906
-0.03559
-0.005896
0.012538
0.014233
0.010571
0.016676
0.003775
0.002331
0.001365
-0.002513
-0.00344
0.014798
country
THA
ETH
TUR
HKG
IDN
IND
TWN
MYS
CRI
MEX
PHL
SGP
BRA
COL
PER
ARG
CHL
KOR
BOL
VEN
region
ASIA
AFRICA
TURKEY
ASIA
ASIA
ASIA
ASIA
ASIA
LAC
LAC
ASIA
ASIA
LAC
LAC
LAC
LAC
LAC
ASIA
LAC
LAC
total
0.030511
0.018733
0.031586
0.03272
0.027799
0.042316
0.039907
0.040816
0.0125
0.01067
0.009455
0.037079
0.004444
0.001849
0.034072
0.023534
0.029252
0.038972
0.008808
-0.003542
within
0.013835
0.003935
0.017353
0.020182
0.017228
0.032409
0.03448
0.035871
0.008725
0.008339
0.00809
0.037853
0.006957
0.00529
0.038536
0.029429
0.038205
0.052947
0.033657
0.032048
structural
0.016676
0.014798
0.014233
0.012538
0.010571
0.009906
0.005427
0.004946
0.003775
0.002331
0.001365
-0.000775
-0.002513
-0.00344
-0.004465
-0.005896
-0.008953
-0.013975
-0.024849
-0.03559
Looking closer at “structural change”
term: LAC
 = -2.6866; t-stat = -1.17
min pu
min
pu
pu
min
minmin
pu
pu
pu
min
pu
minfire
tsc tsc
fire
tscfire
man
con
man
man
man
man man
man tsc
putsctsc man
tsc
mancon
cspsgs
fire
tsc
tsc
fire
con
con
w rt
con
w
rt
cspsgs agr cspsgs
fire
fire
w rt fire cspsgs
cspsgs
w rt
con cspsgs
cspsgs
cspsgs
agragr
cspsgs
fire
w
rt
agr min con
w rt
w rt
agr
agr
pu
rt
agr wcon
con
0
1
2
3
Correlation Between Sectoral Productivity and
Change in Employment Shares in Latin America (1990-2005)
agr
-1
agr
-.2
-.1
0
Change in Employment Share
(Emp. Share)
Fitted values
*Note:  denotes coeff. of independent variable in regression equation:
ln(p/P) =  + Emp. Share
Source: Authors' calculations with data from Timmer and de Vries (2007)
.1
w rt
.2
Looking closer at “structural change”
term: Asia
 = 3.3202; t-stat = 2.04
1
2
3
Correlation Between Sectoral Productivity and
Change in Employment Shares in Asia (1990-2005)
man
0
man
-1
man
agr
agr
agr
agr
-2
agr
agr
agr
man
min
pu
pu
pu
min
min
fire
pu
pu
pu
min
pu fire
min
tsc
min
tsc
min
pu
man
fire
man
pu w rt
man
fire
con
tsc
fire
cspsgs
man fire
fire
tsc tsc
mancspsgs
fire
con
tsctsctsccspsgs
min
tsc
w rt con con
rt rt w rt
wwrtw
w rt
cspsgs
cspsgs
w rt
cspsgs
cspsgs
cspsgs
cspsgs
min
con con w rt
con
con
con
agr
fire
agr
-.2
-.1
0
Change in Employment Share
(Emp. Share)
Fitted values
*Note:  denotes coeff. of independent variable in regression equation:
ln(p/P) =  + Emp. Share
Source: Authors' calculations with data from Timmer and de Vries (2007)
.1
Selected countries: Argentina
2
Correlation Between Sectoral Productivity and
Change in Employment Shares in Argentina (1990-2005)
 = -7.0981; t-stat = -1.21
min
man
tsc
0
.5
1
1.5
pu
con
agr
-.5
w rt
fire
-.06
-.04
-.02
0
Change in Employment Share
(Emp. Share)
Fitted values
*Note: Size of circle represents employment share in 1990
**Note:  denotes coeff. of independent variable in regression equation:
ln(p/P) =  + Emp. Share
Source: Author's calculations with data from Timmer and de Vries (2007)
.02
cspsgs
.04
Selected countries: Brazil
Correlation Between Sectoral Productivity and
Change in Employment Shares in Brazil (1990-2005)
 = -2.2102; t-stat = -0.17
2
pu
1
min
fire
man
tsc
0
con
cspsgs
-1
agr
w rt
-.1
-.05
0
Change in Employment Share
(Emp. Share)
Fitted values
*Note: Size of circle represents employment share in 1990
**Note:  denotes coeff. of independent variable in regression equation:
ln(p/P) =  + Emp. Share
Source: Author's calculations with data from Timmer and de Vries (2007)
.05
Selected countries: India
2
Correlation Between Sectoral Productivity and
Change in Employment Shares in India (1990-2005)
 = 35.2372; t-stat = 2.97
pu
fire
1
tsc
min
wcon
rt
man
-1
0
cspsgs
agr
-.04
-.02
0
Change in Employment Share
(Emp. Share)
Fitted values
*Note: Size of circle represents employment share in 1990
**Note:  denotes coeff. of independent variable in regression equation:
ln(p/P) =  + Emp. Share
Source: Author's calculations with data from Timmer and de Vries (2007)
.02
Selected countries: Thailand
3
Correlation Between Sectoral Productivity and
Change in Employment Shares in Thailand (1990-2005)
 = 5.1686; t-stat = 1.27
pu
2
min
1
tsc
man
0
fire
cspsgs
w rt
-1
con
agr
-.2
-.1
0
Change in Employment Share
(Emp. Share)
Fitted values
*Note: Size of circle represents employment share in 1990
**Note:  denotes coeff. of independent variable in regression equation:
ln(p/P) =  + Emp. Share
Source: Author's calculations with data from Timmer and de Vries (2007)
.1
How does Turkey compare?
Decomposition of productivity growth, Turkey
1988-2008
within
structural change
1990-2005
0.00%
0.50%
1.00%
1.50%
2.00%
2.50%
3.00%
3.50%
Selected countries: Turkey
Correlation Between Sectoral Productivity and
Change in Employment Shares in Turkey (1988-2008)
pu
tsc
.5
1
1.5
 = 2.9138; t-stat = 1.14
fire
min
man
0
con
-.5
w rt
-1
cspsgs
agr
-.2
-.1
0
Change in Employment Share
(Emp. Share)
Fitted values
*Note: Size of circle represents employment share in 1988
**Note:  denotes coeff. of independent variable in regression equation:
ln(p/P) =  + Emp. Share
Source: Authors' calculations with data from the Turkish Statistical Institute
.1
Some intermediate conclusions



Structural change in LAC contributed negatively (unweighted) or
very little (weighted) to labor productivity growth compared to what
happened in ASIA
These economies are supposed to have become more “open”: yet in
all cases the employment share of tradables (the sectors
experiencing the most rapid productivity growth) has been shrinking.
 Some of this is normal, and associated with increase in incomes
 But also signs that some of it is pre-mature
Turkey looks decidedly more “Asian” in terms of the contribution of
structural change to overall productivity growth
Explaining differences across countries



Richer countries may have less room for productivityenhancing structural change
Labor market rigidities may prevent expansion of more
productive sectors
Trade/industrial/currency policies may:



play a role in encouraging/discouraging new tradable activities
expose tradables to import competition too early and excessively
Comparative advantage may

encourage specialization in primary products instead of
manufacturing

with limited potential to absorb labor
Income differences
Average economy-wide labor productivity, 2000 PPP $
60000
1990
2005
50000
40000
30000
20000
10000
0
ASIA
HI
LAC
TURKEY
Inter-sectoral productivity gaps
Dispersion of sectoral labor productivity, 2005
(coefficient of variation of log sectoral productivites)
0.16
Asia average: 0.098
LAC average: 0.098
0.14
Turkey: 0.080
0.12
0.1
0.08
0.06
0.04
0.02
0
HKG
IDN
IND
KOR
MYS
PHL
SGP
THA
TWN
ARG
BOL
BRA
CHL
COL
CRI
MEX
PER
VEN
TUR
.01
Structure of exports
TUR
0
HKG
JPN
THA
FRA
ITASWEESP
MEX
DNKNLD
CRI IDN
-.01
IND
MYS
BRA
PER
SGP
UKM
USA
PHL
-.02
ARG
COL
KOR
-.03
CHL
BOL
-.04
t-stat: -3.38
0
20
VEN
40
Exp_rawmat
60
80
.02
The real exchange rate
t-stat: 3.12
IDN
.01
THA
TUR
PER
CRI
IND
0
BRA
ARG
MEX
HKG
-.01
DNK
SWE
FRA ESP
NLD
ITA
COL
CHL
MYS
PHL
BOL
SGP
JPN
-.02
UKM
USA
VEN
KOR
-1
-.5
0
underval
.5
1
.01
Rigidity of labor laws
TUR
THA
PER
IDN
CRI
DNK
0
HKG
ARG
COL
MYS
MEX
NLD
FRA
SWE
BRAESP
ITA
CHL
JPN
IND
-.01
SGP
USA
UKM
PHL
-.02
BOL
KOR
VEN
-.03
t-stat: -1.70
0
20
40
60
Employment rigidity index (0=less rigid, 100=more rigid)
80
Concluding comments



Structural transformation is the key to economic
development
Structural transformation is not an automatic
process
We need to avoid both the dirigiste and the market
fundamentalist traps



pragmatism rather than preconceived ideologies
strategic collaboration between government and the private
sector
Ingredients:

Institutionalized dialogues, carrots, and sticks