Embed Size (px)
Citation preview
Structural Transformation in India:The Role of the Service Sector
Rafael Serrano-QuinteroUniversity of Alicante
STEG Annual Conference 2021
Introduction
Why India? Major dierences in sectoral labor productivity growth.
Standard experience: Agriculture > Manufacturing > Services.
Indian experience: Services > Agriculture > Manufacturing (Broadberry andGupta, 2010).
Why?
Labor Productivity Trends19
81
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
-1
-0.5
0
0.5
1
1.5
2
2.5
Lo
g o
f R
eal
Lab
or
Pro
du
ctiv
ity
High Manufacturing
Low Manufacturing
High Services
Low Services
Figure 1: Labor Productivity in India
Classify industries inHigh Productivity(HP) vs LowProductivity (LP)growth (Duerneckeret al., 2019).
High-services >High-manufacturing
Low-services >Low-manufacturing
Model
Accounting-type of model (Buera et al., 2020; Herrendorf and Fang, 2019).
Sequence of static economies.
Five sectors of production. Agriculture, HP and LP Manufacturing, HP and LP Services.
High and low skill labor.
Exogenous series: TFPs, Skill-biased technical change, supply of skilled workers, sectoral
distortions.
Calibrated Exogenous Series
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
-0.5
0
0.5
1
1.5
2
2.5TFPs
Agriculture
High Manufacturing
Low Manufacturing
High Services
Low Services
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
0
0.2
0.4
0.6
0.8High-Skill Relative Weight
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
0
1
2
3
4
5
6
7Distortions
High Manufacturing
Low Manufacturing
High Services
Low Services
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
0
0.05
0.1
0.15
0.2Relative Supply of High-Skill Workers
Data
Model
Figure 2: Calibrated Parameters
Conclusions TFPs, though the main driver of growth, cannot explain dierences across
HP-sectors.
Holding the supply of skilled workers constant at 1981 values: Reduces GDP by half.
Removing distortions imply dierent gains on GDP depending on the sector. GDP ×1.5 if HP-Manufacturing.
GDP ×2.6 if HP-Services.
Services-led growth is crucially aected by the supply of skilled workers andwithout a continuous expansion, premature deindustrialization might blockthe road towards economic convergence.
Returns to Schooling
0
.5
1
1.5
2
Estim
ated
Log
-Wag
e
0 4 8 12 16Years of Schooling
1983
0
.5
1
1.5
2
Estim
ated
Log
-Wag
e
0 4 8 12 16Years of Schooling
1987
0
.5
1
1.5
2
Estim
ated
Log
-Wag
e
0 4 8 12 16Years of Schooling
1993
0
.5
1
1.5
2
Estim
ated
Log
-Wag
e
0 4 8 12 16Years of Schooling
1999
0
.5
1
1.5
2
2.5
Estim
ated
Log
-Wag
e
0 4 8 12 16Years of Schooling
2004
Agriculture
High Manufacturing
Low Manufacturing
High Services
Low Services
Figure 3: Log of Wages and Returns to Schooling by Sector
Framework
Discrete time. Sequence of static economies.
Five sectors of production. Agriculture, HP and LP manufacturing, and HP and LP services.
A proportion Mht of household members are high-skilled.
A proportion Mlt of household members are low-skilled.
Each sector employs both high-skilled and low-skilled labor.
Sectoral labor market distortions and free mobility of labor.
Households
Household’s Problem:
U = log(Ct) (1a)
Ct =[(ωa)
1ε (cat)
ε−1ε + (ωm)
1ε (cmt)
ε−1ε + (ωs)
1ε (cst)
ε−1ε
] εε−1 (1b)
cmt =
[(ωh
m
) 1ηm(
chmt
) ηm−1ηm
+(
1−ωhm
) 1ηm(
clmt
) ηm−1ηm
] ηmηm−1
(1c)
cst =
[(ωh
s
) 1ηs(
chst
) ηs−1ηs
+(
1−ωhs
) 1ηs(
clst
) ηs−1ηs
] ηsηs−1
(1d)
patcat + phmtc
hmt + pl
mtclmt + ph
stchst + pl
stclst = wh
t Mht + wlt Mlt + Tt (1e)
Price Eect Long-Run Income Eects
Firms
Firms’ Problem: Alternative PF
maxhi
jt,lijt
pijtY
ijt − (1 + τi
jt)(wht hi
jt + wljtl
ijt) (2a)
s.t. Yijt = Ai
jtLijt = Ai
jt
[πi
jt
(hi
jt
) σ−1σ
+ (1− πijt)(
lijt
) σ−1σ
] σσ−1
(2b)
wht
wlt=
πijt
1− πijt
(lijt
hijt
) 1σ
(3)
Equilibrium
Skilled wages bill:
Ωij ≡
whhij
whhij + wl li
j=
(1 +
(wh
wl
)σ−1(1− πij
πij
)σ)−1
(4)
High-skilled workers over labor input:
hij
Lij=
(πi
j
Ωij
) σ1−σ
(5)
Relative Prices:pi
j
pa=
Aa
Aij
(1 + τi
j
1 + τa
)(πa
πij
) σσ−1(
Ωa
Ωij
) 11−σ
(6)
EquilibriumRelative skill intensity:
hij
ha= Ei
ja
(1 + τa
1 + τij
)(Ωi
j
Ωa
)(7)
Skilled workers in agriculture:
ha
Mh=
1
∑j∈a,m,s
∑i∈h,l
Eija
(1 + τa
1 + τij
)(Ωi
j
Ωa
) (8)
Real labor productivity:
Yij
lij + hi
j=
1
1 +(
wh
wl
)σ(
1−πij
πij
)σ Aij
(πi
j
Ωij
) σσ−1
(9)
Alternative Utility Function Function
Consumption aggregator Ct defined implicitly (Comin et al., 2020) by
ω1/εa
(cat
Cνat
) ε−1ε
+ ω1/εm
(cmt
Cνmt
) ε−1ε
+ ω1/εs
(cst
Cνst
) ε−1ε
= 1 (10)
FOCs give relative demands:
pjcj
paca=
ωj
ωa
(pj
pa
)1−ε
C(1−ε)(νj−νa) (11)
Back
Alternative Production Function
Yijt =
[πi
j
(Λi
jthijt
) σ−1σ
+ (1− πij)(
Γijtl
ijt
) σ−1σ
] σσ−1
Λijt ≡ High-skill augmenting technical change.
Γijt ≡ Low-skill augmenting technical change.
πijt and Ai
jt can be expressed as functions of Λijt, Γi
jt, and σ. Back
πijt ≡
πij
(Λi
jt
) σ−1σ
πij
(Λi
jt
) σ−1σ
+(1−πij)(
Γijt
) σ−1σ
; Aijt ≡
(πi
j
(Λi
jt
) σ−1σ
+ (1− πij)(
Γijt
) σ−1σ
) σσ−1
Households
phjtc
hjt
pljtc
ljt=
(ωh
j
1−ωhj
)(ph
jt
pljt
)1−ηj
for j ∈ m, s (12)
Price eect only.
Within services: HP rising relative to LP.
Relative price is declining.
Elasticity ηs > 1
Within manufacturing
HP declining relative to LP.
Relative price declining.
Elasticity ηm < 1.
Back
Calibration Results
1985 1990 1995 2000 2005 2010 20153
3.5
4
4.5
5Skill Premium
Data
Model
1985 1990 1995 2000 2005 2010 20150
2
4
6
8
10
12
14Relative Nominal Labor Productivities
1985 1990 1995 2000 2005 2010 20150.4
0.6
0.8
1
1.2
1.4Relative Prices
High-Man - Data
High-Man - Model
Low-Man - Data
Low-Man - Model
High-Ser - Data
High-Ser - Model
Low-Ser - Data
Low-Ser - Model
1985 1990 1995 2000 2005 2010 20150
0.05
0.1
0.15
0.2High-Low Skill Labor Ratios
Agriculture - Data
Agriculture - Model
High-Man - Data
High-Man - Model
Low-Man - Data
Low-Man - Model
1985 1990 1995 2000 2005 2010 20150
0.1
0.2
0.3
0.4
0.5
0.6High-Low Skill Labor Ratios
High-Ser - Data
High-Ser - Model
Low-Ser - Data
Low-Ser - Model
1985 1990 1995 2000 2005 2010 20150
5
10
15
20
25
30Aggregate per capita GDP
Figure 4: Targeted Variables
Supply of Skilled Workers
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2
4
6
8
10
12
14
16
Skill Premium
(a)
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8Real Aggregate per capita GDP
Benchmark
Exper
(b)
Figure 5: Counterfactual: Constant Mh/Ml
Additional Tables
Table 1: Division of Manufacturing by Labor Productivity GrowthHigh Productivity Manufacturing
Coke, Refined Petroleum Products and Nuclear fuel 6.0102Chemicals and Chemical Products 5.5726Textiles, Textile Products, Leather and Footwear 4.9469Transport Equipment 4.8449Other Non-Metallic Mineral Products 4.5473Electricity, Gas and Water Supply 4.4339Rubber and Plastic Products 4.0767Manufacturing, nec; recycling 3.4973Food Products,Beverages and Tobacco 3.0075Pulp, Paper,Paper products,Printing and Publishing 2.8913Mining and Quarrying 2.4697Electrical and Optical Equipment 1.8890Basic Metals and Fabricated Metal Products 1.6030
Overall Manufacturing Sector 1.3643
Low Productivity Manufacturing
Machinery, nec. 1.1631Wood and Products of wood -0.5881Construction -1.9478Note: All numbers are in percentages. Labor productivity is the ratio of real value added to quality-
adjusted labor, the numbers represent averages for the full period (1981-2016). Overall ManufacturingSector represents the growth rate of labor productivity in the aggregated manufacturing sector.
What are the distortions?
Female labor market participation (Ngai and Petrongolo, 2017).
Peak in 2005 at 31.8%, 20.5% in 2019 (World Bank, WDI).
Table 2: Distribution of Female Employment in India
1983 1987 1993 1999 2004 2009
Non-Services 86.89 87.20 85.55 84.73 83.25 82.56High services 1.63 1.86 2.48 2.21 2.25 2.94Low services 11.48 10.94 11.97 13.06 14.50 14.50
Migration costs and educational complementarities.
What are the distortions?
Migration costs limit structural change (Alonso-Carrera and Raurich, 2018).
Estimate the following regression
Nsd,t = α + β1 log(Cityd) + β2 log(Railroadd) + β3Sd,t + β4Sd,t × log(Cityd)
+ γ1Longituded + γ2Latituded + µt + εd,t
(13)
Nsd,t ≡ sector s employment share
in district d at time t.
Cityd ≡ distance from district d tothe closest city with more than 1million inhabitants.
Railroadd ≡ distance to closestrailroad.
Sd,t ≡ average years of schooling indistrict d at time t.
What are the distortions?
Figure 6: Distributions of Large Cities and Railroads
What are the distortions?
Table 3: Employment Shares and Distance to Railroads, and Large Cities
Agriculture HighManufacturing
LowManufacturing
HighServices
LowServices
(1) (2) (3) (4) (5)
Distance to Large Cities (logs) 0.034*** -0.019*** 0.004 -0.004** -0.015***(0.005) (0.003) (0.003) (0.002) (0.002)
Distance to Rails (logs) 0.002 -0.007*** 0.003* 0.005*** -0.002**(0.003) (0.001) (0.002) (0.001) (0.001)
Average years of schooling -0.067*** 0.011*** 0.002 0.028*** 0.027***(0.003) (0.001) (0.001) (0.002) (0.001)
City × School 0.013*** -0.001 -0.004*** -0.004** -0.005***(0.002) (0.001) (0.001) (0.002) (0.001)
Controls Yes Yes Yes Yes YesObservations 1648 1648 1648 1648 1648R2 0.482 0.273 0.170 0.410 0.449Data: IPUMS-I. Robust standard errors in parenthesis. ∗ p < 0.10, ∗∗ p < 0.05, ∗∗∗ p < 0.01. Controls include longitude,
latitude, and year fixed eects. Large cities are those cities with more than one million inhabitants. The interaction term iscomputed by first de-meaning each of the variables and then computing the product. Distance to large cities is computedas the minimum distance from the centroid of the district to all cities with more than 1 million inhabitants.
Additional Tables
Table 4: Division of Services by Labor Productivity Growth
High Productivity Services
Post and Telecommunication 8.5416Public Administration and Defense;Compulsory Social Security 4.6582
Business Service 3.9885Financial Services 3.9528
Overall Service Sector 3.6198
Low Productivity Services
Trade 3.4951Health and Social Work 2.9357Education 2.8785Hotels and Restaurants 2.7428Transport and Storage 2.1658Other services 1.3643Note: All numbers are in percentages (%). Labor productivity is the ratio of real value added to
quality-adjusted labor, the numbers represent averages for the full period (1981-2016). OverallService Sector represents the growth rate of labor productivity in the aggregated service sector.
Cross-Country ComparisonsEstimate:
log(LPs,c,t) = α + β1 log(yc,t) + β2(log(yc,t))2 + β3 log(popc,t)
+ϕc + φTimet + γTimet × INDc,t + εs,c,t
LPs,c,t is labor productivity in sectors, country c, at time t.
yc,t is GDP per capita of country cat time t.
popc,t is population of country c attime t.
ϕc denotes country fixed eect.
Timet is a time trend.
Timet × INDc,t is the time trendinteracted with a dummy variablefor India. Coecient of interest.
Cross-country Comparisons
Table 5: Cross-country Comparison of Labor Productivity Growth
(1) (2) (3)Agriculture Manufacturing Services
Time × India -0.0123∗∗∗ -0.00337 0.0177∗∗∗(0.000941) (0.00271) (0.00153)
Time 0.0400∗∗∗ 0.0182∗∗∗ -0.00277(0.00136) (0.00203) (0.00143)
Log of GDP per capita -0.454∗ 2.559∗∗∗ 0.814∗∗∗(0.204 ) (0.414 ) (0.240 )
Log of GDP per capita squared 0.0433∗∗∗ -0.106∗∗∗ -0.0149(0.0121) (0.0236) (0.0135)
Log of Population -1.139∗∗∗ -1.014∗∗∗ -0.175∗∗(0.0595) (0.0812) (0.0660)
Country Fixed Eects Yes Yes YesNo. Countries 41 41 41N 2158 2168 2168
Data: GGDC 10-Sector Database and Maddison Project Database. Robust standard errors inparenthesis. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001.
Cross-Country Comparisons
Robustness:
Within Asia. Asian Countries
Within Asia excludingChina. Excluding China
Using GDP per capita from PennWorld Tables. PWT
Using World DevelopmentIndicators Database (≈145countries). WDI
Within low-income countries(WDI). Low-Income
By region. Regions
Additional Tables
Table 6: Labor Productivity in India Within Asia
(1) (2) (3)Agriculture Manufacturing Services
Time × India -0.00779*** -0.0161*** 0.00665***(0.000768) (0.00392) (0.00141)
Time 0.0122*** 0.0293*** 0.0131***(0.00276) (0.00411) (0.00187)
Log of GDP per capita 0.977*** 2.693*** 0.573***(0.180) (0.371) (0.165)
Log of GDP per capita squared -0.0238* -0.116*** -0.00595(0.00973) (0.0229) (0.00915)
Log of Population -0.582*** -0.911*** -0.266**(0.110) (0.110) (0.0813)
Country Fixed Eects Yes Yes YesNo. Countries 11 11 11N 520 522 522
Data: GGDC 10-Sector Database and Maddison Project Database. Robust standard errors inparenthesis. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001.
Back
Additional Tables
Table 7: Labor Productivity in India Within Asia Excluding China
(1) (2) (3)Agriculture Manufacturing Services
Time × India -0.0110*** 0.0101*** 0.0153***(0.00109) (0.00190) (0.00145)
Time 0.0150*** 0.00799*** 0.00689***(0.00270) (0.00182) (0.00186)
Log of GDP per capita 1.363*** -0.778** -0.798***(0.273) (0.259) (0.201)
Log of GDP per capita squared -0.0471** 0.0872*** 0.0711***(0.0150) (0.0146) (0.0108)
Log of Population -0.612*** -0.576*** -0.0864(0.111) (0.0825) (0.0921)
Country Fixed Eects Yes Yes YesNo. Countries 10 10 10N 461 462 462
Data: GGDC 10-Sector Database and Maddison Project Database. Robust standard errors inparenthesis. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001.
Back
Additional Tables
Table 8: Cross-country Comparison of Labor Productivity Growth
(1) (2) (3)Agriculture Manufacturing Services
Time × India -0.0114*** 0.0139*** 0.0265***(0.000713) (0.00173) (0.00123)
Time 0.0429*** 0.0156*** -0.00255(0.00141) (0.00235) (0.00160)
Log of GDP per capita -0.292* -0.493 -0.502*(0.136) (0.311) (0.195)
Log of GDP per capita squared 0.0285*** 0.0508** 0.0488***(0.00857) (0.0182) (0.0116)
Log of Population -1.188*** -0.670*** -0.0313(0.0592) (0.0855) (0.0657)
Country Fixed Eects Yes Yes YesNo. Countries 41 41 41N 2158 2168 2168
Data: GGDC 10-Sector Database, Maddison Project Database, and Penn World Tables. Robuststandard errors in parenthesis. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001.
Back
Additional Tables
Table 9: Labor Productivity in India
(1) (2) (3)Agriculture Manufacturing Services
Time × India -0.00714*** -0.00396** 0.0219***(0.00134) (0.00143) (0.000945)
Time 0.0191*** -0.00261* -0.00154**(0.00191) (0.00105) (0.000596)
Log of GDP per capita 1.711*** -0.471** 0.465***(0.272) (0.178) (0.107)
Log of GDP per capita squared -0.0671*** 0.0752*** 0.00934(0.0171) (0.0101) (0.00589)
Log of Population -0.875*** -0.0901 -0.0666*(0.0818) (0.0525) (0.0299)
Country Fixed Eects Yes Yes YesN 3681 3671 3504
Data: World Development Indicators. These regressions exclude oil-exporting countries asclassified by the IMF. Robust standard errors in parenthesis. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗p < 0.001.
Back
Additional Tables
Table 10: Labor Productivity in India (only Low-Income Countries)
(1) (2) (3)Agriculture Manufacturing Services
Time × India -0.00315 -0.000482 0.0216***(0.00186) (0.00167) (0.00120)
Time -0.0000983 -0.0108*** 0.00341(0.00341) (0.00239) (0.00189)
Log of GDP per capita 2.191*** -0.339 0.0979(0.313) (0.316) (0.261)
Log of GDP per capita squared -0.0897*** 0.0644** 0.0268(0.0213) (0.0195) (0.0165)
Log of Population -0.237 0.354*** -0.118(0.127) (0.104) (0.0728)
Country Fixed Eects Yes Yes YesN 1275 1261 1196
Data: World Development Indicators. These regressions exclude oil-exporting countries asclassified by the IMF. Robust standard errors in parenthesis. ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗p < 0.001.Regressions include only those countries considered as low-income countries bythe World Bank in the year 2000.
Back
Additional Tables
Table 11: Dierential Labor Productivity Growth by RegionAgriculture Manufacturing Services
Panel A: Africa
Time × Region -0.00721*** -0.00116 0.0121***(0.000964) (0.00134) (0.00135)
Time 0.0391*** 0.0179*** -0.00159(0.00130) (0.00190) (0.00145)
Panel B: Asia
Time × Region -0.0174*** 0.0173*** 0.0207***(0.000934) (0.00174) (0.00130)
Time 0.0380*** 0.0189*** -0.000165(0.00119) (0.00172) (0.00122)
Panel C: Latin America
Time × Region 0.00813*** -0.00548*** -0.0159***(0.000688) (0.00117) (0.00129)
Time 0.0378*** 0.0186*** 0.000779(0.00126) (0.00194) (0.00133)
Panel D: Western Countries
Time × Region 0.0173*** -0.00900*** -0.0112***(0.00121) (0.00142) (0.00138)
Time 0.0235*** 0.0260*** 0.00876***(0.00165) (0.00234) (0.00196)
Data: GGDC 10-Sector Database and Maddison Project Database. Robust standard errors in parenthesis. ∗ p < 0.05,∗∗ p < 0.01, ∗∗∗ p < 0.001. Each panel shows the result of a separate regression in which the dummy variable Regiontakes value equal to one if the region corresponds to that of the panel and zero otherwise. All regressions includecountry fixed eects and control for log of GDP per capita, log of GDP per capita squared, and population
Back
