The Agricultural Productivity Gap

The Agricultural Productivity Gap

Douglas Gollin David Lagakos Michael E. Waugh

Oxford UCSD NYU

November 11, 2013

0 / 37

Agriculture Sector Across Countries

• Share of value added lower than share of employment

• True in basically every country in world

• Particularly so developing countries

1 / 37

Agriculture Sector Across Countries

0 10 20 30 40 50 60 70 80 90 1000

10

20

30

40

50

60

70

80

90

100

ALB

DZA

ATG

ARG

ARM

AUSAUT

AZEBHS

BGD

BRB

BLR

BEL

BLZ

BEN

BMU

BTN

BOL

BWABRA

BRN

BGR

BFA

BDI

KHM

CMR

CAN

CPV

CAF

TCD

CHL

CHN

COLCRI

CIV

HRV CUBCYPCZEDNK

DMA

DOM

EGYSLV

EST

ETH

FJI

FINFRAGAB

GMB

GEO

GER

GHA

GRC

GRD

GTM

GINGUY

HTI

HND

HUNISL

IND

IDN

IRN

IRQIRLISRITA

JAM

JPNJOR

KAZ

KEN

KOR

KGZ

LAO

LVALBN

LSO

LBR

LTU

MKD

MDGMWI

MYS

MDV

MLI

MLT

MHL

MUS MEX

MDA

MNG

MNEMAR

NAM

NPL

NLDNZL

NIC

NGA

NOR OMN

PAK

PAN

PNG

PRY

PHL

POLPRT

ROMRUS

RWA

LCA

WSM

STP

SAU

SEN

SRB

SLE

SVKSVNZAFESP

LKA

VCT

SDN

SUR

SWZ

SWECHE

SYRTJK

TZA

THA

TON

TUNTUR

UGA

UKR

GBRUSA

URY

UZB

VEN

VNM

YEM

ZMBZWE

Share of Employment in Agriculture

Sh

are

of V

alu

e A

dd

ed

in A

gricu

lture

2 / 37


We define the Agricultural Productivity Gap (APG) to be

APG ≡VAn/Ln

VAa/La

.

• Simple two-sector model says APG should be 1

• In practice, average APG ∼ 3

• Poorest quartile of income distribution: average APG = 5.6 !

3 / 37


• Taken at face value, gaps suggest misallocation

• Policy debate: encourage movement out of agriculture?

• This paper: step back and address measurement issues

4 / 37

What Do Agricultural Productivity Gaps Reflect?

• Sector differences in hours worked per worker?

Construct measures of hours worked by sector for 76 countries

• Sector differences in human capital per worker?

Construct measures of human capital by sector for 124 countries

• Measurement error in national accounts data?

Construct our own estimates using household survey data in 10 countries

5 / 37








5 / 37








5 / 37








5 / 37

What We Conclude

• Our adjustments reduce average APG to around two

• Still bigger in developing countries

• Large gaps also present in household survey data

• Needed: better understanding of why residual gaps so large

6 / 37

Simple Two-Sector Model

• Technologies

Ya = AaLθaK

1−θa and Yn = AnL

θnK

1−θn

• Households can supply labor to either sector.

• Competitive labor markets, i.e. workers paid their marginal product.

• Equilibrium:

APG ≡VAn/Ln

VAa/La

=Yn/Ln

paYa/La

= 1.

7 / 37

Computing “Raw” Agricultural Productivity Gaps

Measures of VAa and VAn

• Value added as defined in 1993 System of National Accounts (SNA)

• Source: UN National Account Statistics

Measures of La and Ln

• Employed persons working in the production of some good or service

recognized by the 1993 SNA

• Source: ILO, via population censuses or labor force surveys.

8 / 37

Raw Agricultural Productivity Gaps

Quartile of Income Distribution

All Countries Q1 Q2 Q3 Q4

10th Percentile 1.3 1.0 1.3 1.0 1.2

Median 2.6 1.7 2.7 2.8 4.3

Mean 3.5 2.0 3.2 3.4 5.6

90th Percentile 6.8 4.0 6.6 7.1 12.5

Number of Countries 151 38 38 38 37

9 / 37

“Simple” Measurement Error in National Accounts Data?

1. Understate agricultural VA by excluding home production?

• In principal: No, it is included as per SNA.

• Accepted practice: output of particular crop = area planted X yield

2. Overstate agricultural employment, by including all rural persons?

• In principal: No, only economically active persons included per SNA.

• We find national accounts consistent with household surveys.

10 / 37

Our Adjustments

1. Improved measures of labor input by sector

• Sector differences in hours worked per worker.

• Sector differences in human capital per worker.

2. Alternative measures of value added by sector

• Reconstruct national accounts data from household survey data.

11 / 37

Our Adjustments

1. Improved measures of labor input by sector

• Sector differences in hours worked per worker.

• Sector differences in human capital per worker.

2. Alternative measures of value added by sector

• Reconstruct national accounts data from household survey data.

11 / 37

Improved Measures of Labor Input by Sector

12 / 37

Sector Differences in Hours Worked

Average hours worked per worker might differ across sectors

We construct average hours worked per worker by sector for 76 countries

• Population census micro data or labor force surveys

• All employed persons 15+ years old

• Industry of primary employment

• Hours worked in reference period (usually one week)

13 / 37


25 30 35 40 45 50 5525

30

35

40

45

50

55

ALB

ARM

AUS

BGD

BEL

BTN

BOL

BWA

BRA

BGR

KHM

CAN

CHL

CIV

EGY

SLV

ESTETH

FJI

FIN FRAGER

GHA

GRC

GTM

HND

HUN

IDN

IRQ

ISR ITA

JAM

JORKEN

KOR

LVA

LSO

LBR

LTU

MWIMUS

MEX

NPL

NLD

NIC

NGA

PAK

PAN

PRY

PHL

POL PRT

ROM

RUS

RWA

LCA

SLE

SVN

ZAF

ESP

LKA

SWZ

SWE CHE

SYR

TJK

TZA

TON

TUR

GBR

USA

VEN

VNM

ZMB

ZWE

Hours Worked in Agriculture

Hou

rs W

orke

d in

Non

−A

gric

ultu

re

2.0 1.5 1.0

14 / 37


Agricultural Productivity Gaps and Hours Adjustment

Measure Raw APG Hours Adjustment

5th Percentile 1.4 1.6

Median 3.2 2.7

Mean 3.1 2.9


Number of Countries 76 76

Note: All statistics are weighted with each country weighted by its population.Only

countries with hours data shown.

Differences in hours worked contribute a factor of 1.1.

15 / 37

Not Artifact of Seasonality

2 4 6 8 10 12

20

30

40

50

Ghana

Ho

urs

Wo

rke

d

2 4 6 8 10 12

20

30

40

50

Malawi

Ho

urs

Wo

rke

d

2 4 6 8 10 12

20

30

40

50

Tanzania

Ho

urs

Wo

rke

d

2 4 6 8 10 12

20

30

40

50

Uganda

Month

Ho

urs

Wo

rke

d

2 4 6 8 10 12

20

30

40

50

Vietnam

Month

Ho

urs

Wo

rke

d

Non−Agriculture Agriculture

16 / 37

Not Artifact of Secondary Jobs

Sector of Hours Worked

Country Worker Classification Agriculture Non-agriculture

Cote d’Ivoire (1988) Agriculture 35.1 1.0

Non-agriculture 0.7 49.2

Ghana (1998) Agriculture 28.8 3.7


Guatemala (2000) Agriculture 47.6 1.3


Malawi (2005) Agriculture 26.4 1.4


Tajikistan (2009) Agriculture 39.5 0.1


Uganda (2009) Agriculture 18.7 2.1


17 / 37

Sector Differences in Human Capital

Average human capital per worker could differ across sectors

(Caselli & Coleman, 2001; Vollrath, 2009; Herrendorf & Schoellman, 2013)

We construct human capital per worker by sector for 124 countries

• Years of schooling measured directly when available

• Impute years of schooling using educational attainment otherwise

• Baseline: Assume 10% rate of return on year of schooling

hj,i = exp(sj,i · 0.10)

18 / 37

Sector Differences in Schooling

0 5 10 150

5

10

15

ALB

ARG

ARM

AUS

AUT

AZE

BGD

BLR

BLZ

BEN

BTN

BOL

BWABRA

BGR

BFA

BDI

KHM

CMR

CAN

CAF

TCD

CHL

CHNCOL CRI

CIV

CUB DNK

DOM

EGY

SLV

EST

ETH

FJI

FRA

GAB

GMB

GEO GER

GHA

GRC

GTM

GIN

GUY

HTI

HND

HUN

ISL

IND

IDN

IRN

IRQ

IRL

ISRITA

JAM

JOR

KAZ

KEN

KOR

KGZ

LAO

LVA

LSO

LBR

LTUMKD

MDG MWI

MYSMDV

MLI

MHL

MEX

MDAMNG

MAR

NAM

NPL

NLD

NIC NGA

NOR

PAK

PAN

PNG

PRYPHL

PRT

ROMRUS

RWA

LCA

STP

SEN

SRB

SLE

SVN

ZAFESP LKA

SDN

SUR

SWZ

SWE

CHE

SYR

TJK

TZA

THA

TON

TUR

UGA

UKR

GBR

USA

URY

UZB

VEN VNM

YEM

ZMB

ZWE

Years of School in Agriculture

Yea

rs o

f Sch

ool i

n N

on−

Agr

icul

ture

2.0 1.5 1.0

19 / 37

Sector Differences in Human Capital

1 1.5 2 2.5 3 3.5 4 4.51

1.5

2

2.5

3

3.5

4

4.5

ALB

ARG

ARM

AUS

AUT

AZE

BGD

BLR

BLZ

BEN

BTN

BOL

BWABRA

BGR

BFA

BDI

KHM

CMR

CAN

CAF

TCD

CHL

CHNCOL

CRI

CIV

CUB DNK

DOM

EGY

SLV

EST

ETH

FJI

FRA

GAB

GMB

GEOGER

GHA

GRC

GTM

GIN

GUY

HTI

HND

HUN

ISL

IND

IDN

IRN

IRQ

IRL

ISRITA

JAM

JOR

KAZ

KEN

KOR

KGZ

LAO

LVA

LSO

LBR

LTUMKD

MDGMWI

MYSMDV

MLI

MHL

MEX

MDA

MNG

MAR

NAM

NPL

NLD

NICNGA

NOR

PAK

PAN

PNG

PRYPHL

PRT

ROM

RUS

RWA

LCA

STP

SEN

SRB

SLE

SVN

ZAF

ESPLKA

SDN

SUR

SWZ

SWE

CHE

SYR

TJK

TZA

THA

TON

TUR

UGA

UKR

GBR

USA

URY

UZB

VEN VNM

YEM

ZMB

ZWE

Human Capital in Agriculture

Hum

an C

apita

l in

Non

−A

gric

ultu

re

2.0 1.5 1.0

20 / 37

Sector Human Capital Differences

Agricultural Productivity Gaps and Human Capital Adjustment

Measure Raw APG Human Capital Adjustment


Median 3.6 2.6

Mean 3.8 2.7


Number of Countries 124 124

Note: All statistics are weighted with each country weighted by its population. Only countries

with schooling data shown.

Human capital differences contribute a factor of 1.4.

21 / 37

Quality Adjustments to Schooling Data

• Rural schools often of lower quality than urban schools

(Williams, 2005; Zhang, 2006)

• Health, parental inputs may be lower in rural areas

• Potentially overestimate human capital among agriculture workers

• We use literacy data to adjust for this

22 / 37

Uganda: Literacy by Years of Schooling Completed

!"

!#$"

!#%"

!#&"

!#'"

("

!" $" %" &" '" (!"

Lit

erac

y R

ate

Years of Schooling

)*)+,-".*/01/2" ,-".*/01/2"

23 / 37

Measuring Quality Differences in Schooling

• Given literacy rates by years of schooling: ℓni (s) and ℓai (s) for s = 1, 2, ...

• Assume that each year in rural school is worth γ years in urban school

• For each country i , solve for γi that solves

minγ

s̄∑

s=1

(

ℓ̃ni (γs)− ℓ̃ai (s))2

where ℓ̃ni (·), ℓ̃ai (·) are polynomial interpolations of ℓni (·), ℓ

ai (·) for s ∈ [0, s̄].

24 / 37

Sector Differences in Quality-Adjusted Human Capital

ARG

BOL

BRA

CHL

GHA

GIN

MYS

MLI

MEX

PAN

PHL

RWA

TZA

THA

UGAVEN VNM

1.01.52.01

23

4

Hum

an C

apita

l in

Non

−A

gric

ultu

re

1 2 3 4

Human Capital in Agriculture

25 / 37

Human Capital: Other Issues

• Country-specific returns to schooling?

- Doesn’t change much

• Quality-adjusted returns to schooling from Schoellman (2012)?

- Modestly decreases importance of human capital

• Human capital from experience?

- Returns somewhat lower in agriculture (Lagakos, Moll, Porzio, Qian,

2013; Herrendorf & Schoellman, 2013)

- Increases importance of human capital

- Limitation: returns estimated for only 20 countries

26 / 37

Adjusting the Raw APG numbers

Recap:

• Differences in hours worked contribute a factor of 1.1.

• Differences in human capital contribute a factor of 1.4.

Now, put them all together and construct “adjusted” APGs.

27 / 37

Adjusted Agricultural Productivity Gaps

Table 4: Agricultural Productivity Gaps and All Adjustments

All Adjustments by Quartile

Measure Raw APG All Adjustments Q1 Q2 Q3 Q4

10th Percentile 1.3 1.0 0.8 1.2 0.7 1.3

Median 3.1 1.9 1.4 2.0 2.1 2.3

Mean 3.5 2.2 1.7 2.1 1.9 3.0

90th Percentile 6.4 4.3 3.3 2.8 4.3 5.6

Number of Countries 72 72 18 16 18 20

28 / 37

Raw vs Adjusted Gaps

2 4 6 8 10 12 14

1

2

3

4

5

6

7

ALB

ARM

AUS

BGD

BTNBOL

BWA

BRA

BGRKHM

CAN

CHL

CIVEGY

SLVEST

ETH

FJI

FRA

GER

GHA

GRC

GTMHND

HUN

IDN

IRQ

ISR

ITA

JAM

JOR

KEN

KOR

LVA LBR

LTU

MWI

MEX

NPL

NLD

NICNGA

PAK

PAN

PRY

PHL

PRT

ROM

RUS

RWA

LCA

SLE

SVN

ZAFESP

LKA

SWZSWE

CHE

SYR

TJK

TZA

TON

TUR

UGA

GBRUSA VEN

VNM

ZMB

ZWE

Raw APG

Ad

just

ed

AP

G

1.0

0.50

29 / 37

Alternative Measures of Value Added by Sector

30 / 37

Comparing Macro and Micro Data on Sector Value Added

The idea:

• Cross check “macro” value added data (from national accounts) with

“micro” data from household income/expenditure surveys.

The data:

• Use World Bank’s Living Standards Measurement Surveys (LSMS)

• Explicit goal of LSMS: household income and expenditure measures

31 / 37

Measuring Value Added from Micro Data

Agriculture:

VAa =∑

i

ySEa,i +

∑

i

yLa,i +

∑

i

yKa,i ,

ySEa,i =

J∑

j=1

pj

(

xhomei,j + x

marketi,j + x

investi,j

)

− COSTSa,i ,

Non-agriculture:

VAn =∑

i

ySEn,i +

∑

i

yLn,i +

∑

i

yKn,i ,

ySEn,i = REVn,i − COSTSn,i .

i = household and j = agriculture commodity.

32 / 37

Comparison of Macro and Micro APG

Agriculture Share of

Country Employment Value Added APG

Micro Macro Micro Macro Micro

Armenia (1996) 34.2 36.8 32.8 0.9 1.1

Bulgaria (2003) 14.1 11.7 18.4 1.2 0.7

Cote d’Ivoire (1988) 74.3 32.0 42.1 4.7 4.0

Guatemala (2000) 40.2 15.1 18.7 3.8 2.9

Ghana (1998) 53.9 36.0 33.3 2.2 2.3

Kyrgyz Republic (1998) 56.9 39.5 39.3 2.0 2.0

Pakistan (2001) 46.9 25.8 22.6 2.5 3.0

Panama (2003) 27.0 7.8 11.8 4.4 2.7

South Africa (1993) 11.0 5.0 7.0 2.3 1.7

Tajikistan (2009) 41.0 24.7 30.1 2.1 1.6

33 / 37





Armenia (1996) 34.2 36.8 32.8 0.9 1.1

Bulgaria (2003) 14.1 11.7 18.4 1.2 0.7

Cote d’Ivoire (1988) 74.3 32.0 42.1 4.7 4.0

Guatemala (2000) 40.2 15.1 18.7 3.8 2.9

Ghana (1998) 53.9 36.0 33.3 2.2 2.3

Kyrgyz Republic (1998) 56.9 39.5 39.3 2.0 2.0

Pakistan (2001) 46.9 25.8 22.6 2.5 3.0

Panama (2003) 27.0 7.8 11.8 4.4 2.7

South Africa (1993) 11.0 5.0 7.0 2.3 1.7

Tajikistan (2009) 41.0 24.7 30.1 2.1 1.6

33 / 37





Armenia (1996) 34.2 36.8 32.8 0.9 1.1

Bulgaria (2003) 14.1 11.7 18.4 1.2 0.7

Cote d’Ivoire (1988) 74.3 32.0 42.1 4.7 4.0

Guatemala (2000) 40.2 15.1 18.7 3.8 2.9

Ghana (1998) 53.9 36.0 33.3 2.2 2.3

Kyrgyz Republic (1998) 56.9 39.5 39.3 2.0 2.0

Pakistan (2001) 46.9 25.8 22.6 2.5 3.0

Panama (2003) 27.0 7.8 11.8 4.4 2.7

South Africa (1993) 11.0 5.0 7.0 2.3 1.7

Tajikistan (2009) 41.0 24.7 30.1 2.1 1.6

33 / 37

Income and Expenditure Per Worker and APGs

Country APG MicroIncome per

Worker Ratio

Expenditure per

Worker Ratio

Armenia (1996) 1.1 0.7 0.9

Bulgaria (2003) 0.7 1.4 1.2

Cote d’Ivoire (1988) 4.0 3.5 3.2

Guatemala (2000) 2.9 3.2 2.4

Ghana (1998) 2.3 2.0 1.9

Kyrgyz Republic (1998) 2.0 1.3 1.8

Pakistan (2001) 3.0 3.2 1.4

Panama (2003) 2.7 2.8 2.1

South Africa (1993) 1.7 1.7 1.2

Tajikistan (2009) 1.6 1.2 1.1

34 / 37

Different Labor Shares Across Sectors?

Production functions with different labor shares

Ya = AaLθaa K

1−θaa and Yn = AnL

θnn K

1−θnn

In equilibrium

APG =Yn/Ln

paYa/La

=θaθn

Macro evidence on θa, θn

• Employment share of agriculture varies a lot across countries;

• Aggregate labor share of GDP doesn’t, Gollin (2002) ⇒ θa ≈ θn

Micro evidence on θa, θn

• Sharecropping arrangements suggest θa ≈ 0.5

• Econometric estimates: θa ≈ 0.5 − 0.6

35 / 37

Why are Residual Gaps So Large?

• Yet more measurement error

– Herrendorf and Schoellman (2013)

• Selection of more productive workers out of agriculture

– Lagakos and Waugh (2013), Young (2013)

• Risk of migrating?

– Harris and Todaro (1971), Bryan, Mobarak, Chowdhury (2012), others

• Much room for future work

36 / 37

Conclusion

• Typical country has large agricultural productivity gap

• Particularly large in developing countries

• Better measurement reduces gap down to around two on average

• Large gaps also present in household survey data

• Needed: better understanding of why residual gaps so large

37 / 37

The Agricultural Productivity Gap

Documents