The (Evolving) Role of Agriculture in Poverty Reduction Luc Christiaensen (UNU-WIDER), Lionel Demery (Development Consultant), Jesper Kuhl (Development.

The (Evolving) Role of The (Evolving) Role of Agriculture in Poverty Agriculture in Poverty

ReductionReduction

Luc Christiaensen (UNU-WIDER), Lionel Demery Luc Christiaensen (UNU-WIDER), Lionel Demery (Development Consultant), Jesper Kuhl (Development Consultant), Jesper Kuhl

(Development Consultant)(Development Consultant)

Presentation at UN-HQ New York, 2 June, 2010Presentation at UN-HQ New York, 2 June, 2010

Four questionsFour questions1)1) Does a focus on agriculture lead to more Does a focus on agriculture lead to more

economic growth than a focus on non-agriculture?economic growth than a focus on non-agriculture?

2)2) Do the poor participate more in growth from Do the poor participate more in growth from agriculture than in growth from non-agriculture?agriculture than in growth from non-agriculture?

3)3) Does potentially greater participation by the poor Does potentially greater participation by the poor in ag growth offset potentially slower growth from in ag growth offset potentially slower growth from ag, and under which circumstances?ag, and under which circumstances?

4)4) Do the results differ depending on the poor Do the results differ depending on the poor groups considered ($1-day vs $2-day)groups considered ($1-day vs $2-day)

FindingsFindings Agriculture better at reducing $1-day poverty; non-Agriculture better at reducing $1-day poverty; non-

agriculture better at reducing $2-day povertyagriculture better at reducing $2-day poverty

High inequality reduces poverty reducing power of High inequality reduces poverty reducing power of agriculture and presence of extractive industries agriculture and presence of extractive industries reduces poverty reducing effects of non-agriculturereduces poverty reducing effects of non-agriculture

Results driven by larger participation of poor in Results driven by larger participation of poor in growth from agriculturegrowth from agriculture

Boosting agricultural performance critical important Boosting agricultural performance critical important to reduce poverty especially in low inc countriesto reduce poverty especially in low inc countries

What follows? What follows? 1)1) A conceptual frameworkA conceptual framework

2)2) Growth potential across sectors–direct growth effects?Growth potential across sectors–direct growth effects?

3)3) Agriculture and the rest of the economy–indirect Agriculture and the rest of the economy–indirect growth effects?growth effects?

4)4) Benefiting from growth – participation effectsBenefiting from growth – participation effects

5)5) The (evolving) role of agriculture in poverty reduction The (evolving) role of agriculture in poverty reduction - a synthesizing perspective- a synthesizing perspective

6)6) Concluding remarksConcluding remarks

Conceptual Conceptual FrameworkFramework

Growth and Poverty reductionGrowth and Poverty reduction

Poverty change = elasticity of poverty to GDP *GDP growthCall the elasticity (ε ), the participation component Two components: a participation and a growth component

dlnPi ≡ εidlnYi

Poverty reducing effect of growth Poverty reducing effect of growth in a sector depends on four in a sector depends on four componentscomponents

Participation Share Growth

pi aisai yai + nisni yni

pit aitsait-1 yait(ynait-k) + nitsnit-1 ynit(yait-k)

participation share direct and indirect growth

• The growth effects depends on 2 components: direct and indirect growth effect

• The elasticity of overall poverty to sectoral GDP depends on 2 components: participation and share component

participation share direct and indirect growth

The Role of Agriculture in Poverty Reduction - schematically

yait

ynait

Pit

aitsait-1

nits nit-1

Growth component

Indirect effectyait(ynait-k) &ynait(yait-k)

Participation and share components

Direct effect of a

Direct effect of n

pit aitsait-1 yait(ynait-k) + nitsnit-1 ynit(yait-k)

The Role of Agriculture in Poverty Reduction – direct growth effect

pit aitsait yait(ynait-k) + nitsnit ynit(yait-k)

yait

ynait

Pit

aitsait

nits nit

Growth component

Indirect effectyait(ynait-k) &ynait(yait-k)

Participation and share component

Direct effect

Direct effect

Growth potential across sectors Growth potential across sectors - direct growth - direct growth effectseffects

AG-a lagging sector incapable of producing rapid growth?

Over past 40 years ag growth has lagged nag growth by about 1.6 percentage points (1.2 in SSA)

Growth = change in productivity + change in input use + change in prices

• Lower growth = lower productivity growth? • Adam Smith: lower growth potential for AG b/c spatial impediments to labour division and capital accumulation

Decompose the growth numbers into their productivity and population growth components

Slower growth in agriculture (1960-2003) largely Slower growth in agriculture (1960-2003) largely associated with migration of workers from AG to associated with migration of workers from AG to

NAG, not because of slower labor productivity NAG, not because of slower labor productivity growthgrowth

Average annual growth Average annual growth rates (1960-2003) rates (1960-2003) (%)(%)

Ag Ag GDPGDP

ProducProductivitytivity

PopuPopuLationLationin agin ag

NonagNonag GDPGDP

ProduProductivityctivity

PopuPopuLationLationIn nagIn nag

Sub-Saharan AfricaSub-Saharan Africa 2.62.6 0.910.91 1.71.7 3.83.8 -0.64-0.64 4.54.5

South Asia South Asia 2.92.9 1.21.2 1.61.6 5.85.8 2.22.2 3.63.6

East Asia & PacificEast Asia & Pacific 2.32.3 2.92.9 -0.5-0.5 5.75.7 2.72.7 2.92.9

Eastern Europe & Central Eastern Europe & Central AsiaAsia

0.810.81 3.43.4 -2.5-2.5 2.62.6 1.41.4 1.21.2

Europe, othersEurope, others 1.51.5 4.64.6 -3-3 2.92.9 22 0.870.87

Latin America & the Latin America & the CaribbeanCaribbean

22 2.32.3 -0.24-0.24 3.33.3 0.480.48 2.82.8

Middle East & North AfricaMiddle East & North Africa 4.44.4 4.34.3 0.210.21 4.74.7 0.260.26 4.44.4

North America North America 1.71.7 3.93.9 -2.1-2.1 33 1.81.8 1.21.2

TotalTotal 2.32.3 2.42.4 -0.05-0.05 3.93.9 0.740.74 3.13.1

Striking antidote to prevailing thinking, but not conclusive

Industrial pull or agricultural Industrial pull or agricultural push?push?

Industrial pull – wage equilibrating labor movements in response to higher marginal productivity outside AG

Agricultural push – productivity growth in ag leads to deterioration of TOT against ag and thus declining remuneration of factors (labor and capital) in agric, inducing migration

Empirical evidence (manufacturing/industrial and TFP) – Szirmai (2009) ag labor productivity growing faster in 12/16

developing countries (Latin America and Asia) bw 1973-2005– from industrial countries suggests TFP growth in ag larger than

in nonag sector (Bernard and Jones,1996)– Recent evidence from developing countries confirms this (Martin

and Mitra, 2001)

Emerging insights on Emerging insights on agriculture’sagriculture’sgrowth potentialgrowth potential No superiority in agricultural TFP growth, but

debunk the notion that agriculture is a backward sector.

Globally AG will grow slower than NAG due to Engel’s Law, not b/c inferior productivity growth.

Important growth opportunities for some countries given high income elasticity for non-staple food and where international trading opportunities exist (Brazil, Chile)

Recovery in ag TFP in SSA over past decade holds promise and expected higher agricultural commodity prices in medium term provide opportunities , though many challenges remain,

The Role of Agriculture in Poverty Reduction - schematically

pit aitsait yait(ynait-k) + nitsnit ynit(yait-k)

yait

ynait

Pit

aitsait

nits nit

Growth components

Indirect effectlaboryait(ynait-k) &ynait(yait-k)

Participation and share components

Direct effect

Direct effect

Production linkages – forward (agro processing) and backward (input

demand, marketing services)

Consumption linkages– increased agric productivity –> direct income effect if

tradables; – indirect effect if non-tradables/food lower food prices,

higher real incomes, higher demand for locally produced goods and services, off-farm employment generation

Real product wage effects – lower food prices, lower real wages in non-agriculture,

higher profits and investment

Agriculture and the rest of the Agriculture and the rest of the economy – indirect growth economy – indirect growth effectseffects

Consumption linkages four times as important as production linkages at early stages of development

Size of consumption linkages depends on:– how broad based growth process; – income propensity for locally produced goods and services;– supply of locally produced goods and services sufficiently elastic.

In SSA, linkage effects situated in the 1.3-1.5 range– one dollar generated in agriculture generates another 30 to 50 cents in the rest

of the economy

Linkages at least as strong from agriculture to nonagriculture than the reverse, with linkage effects declining as countries develop (even though production linkages tend to increase)

Common wisdom so farCommon wisdom so far

New realitiesNew realities

Much of existing evidence is based on Much of existing evidence is based on structural models, and therefore structural models, and therefore questionedquestioned

Increasing urban-rural subcontracting as Increasing urban-rural subcontracting as countries develop countries develop growth in agriculture growth in agriculture no longer engine of growth?no longer engine of growth?

Globalization reduces non-tradability of Globalization reduces non-tradability of food, also in SSA food, also in SSA lower linkage effects? lower linkage effects?

The Tradability of The Tradability of Food?Food?

Reduced form evidence Reduced form evidence Granger Granger causality using GMM dynamic panel causality using GMM dynamic panel estimationestimation

Non-agricultural growth (and vice versa for ag growth):

Granger causality - does lagged ag growth affects current nonag growth?

Estimation: system GMM with share of mining andprecipitation acting as additional external instruments

Growth linkages from AG to NAG Growth linkages from AG to NAG largely limited to low-income SSA largely limited to low-income SSA countriescountries

Nonagricultural growth per capita

Full sampl

e

Middle incom

e

Middle income

Low income

Low income

Non-agr. growth t-1 0.40 0.42 0.40 0.31 0.370.00 0.00 0.00 0.00 0.00

Agr. growth t-1 0.09 0.11 0.11 0.15 0.070.20 0.27 0.30 0.11 0.71

Agr. growtht-1*SSA -0.10 0.120.82 0.58

Mining share t-2 -0.02 -0.09 -0.13 -0.19 -0.120.94 0.69 0.62 0.20 0.30

p-valueAg growtht-1+Ag growtht-1*SSA = 0 0.98 0.09

No linkage effect from NAG to AG in No linkage effect from NAG to AG in SSA, but noticeable reverse linkage SSA, but noticeable reverse linkage effects in other low income effects in other low income countriescountries

Agricultural growth per capita

Full sample

Middle income

Middle income

Low incom

e

Low income

Non-agr. growth t-1 0.07 -0.07 -0.02 0.15 0.180.33 0.52 0.84 0.03 0.01

Non-agr growtht-1*DSSA -0.47 -0.090.00 0.40

Agr. growtht-1 0.21 0.29 0.30 0.12 0.110.01 0.05 0.02 0.10 0.17

Change in precipitation 0.05 0.01 -0.01 0.08 0.090.05 0.81 0.66 0.07 0.04

p-valueNonag growtht-1+ Nonag growtht-1*SSA =0 0.00 0.34

Indirect growth effects – concluding Indirect growth effects – concluding remarksremarks

Evidence from structural and reduced form models suggests:

• Continuing linkage effects from agriculture to non-agriculture in SSA, but no longer when countries develop (see also Henderson, Storeygard and Weil, 2009)

• In SSA these indirect effects are at least as large the reverse linkages, though not in other low income countries (also consistent with the declining share of agriculture in total GDP)

• Variation in exact magnitudes likely depending on agro-ecological conditions, institutions, agrarian structures

The Participation The Participation EffectEffect

yait

ynait

Pit

aitsait

nits nit

Growth components

Indirect effectyait(ynait-k) &ynait(yait-k)

Participation and share components

Direct effect

Direct effect

Three main reasons why poverty Three main reasons why poverty reducing impact differs across reducing impact differs across sectorssectors

Difficulties in transferring income across sectors Difficulties in transferring income across sectors or locations e.g. due to market segmentations or or locations e.g. due to market segmentations or considerations of political economyconsiderations of political economy

Differences in labor intensity across sectorsDifferences in labor intensity across sectors

Differences in inequality in distribution of Differences in inequality in distribution of relevant assets (land in agriculture and capital in relevant assets (land in agriculture and capital in non-agriculture)non-agriculture)

Estimating the participation Estimating the participation effect -effect - – empirical – empirical methodologymethodology

• Test π at=πnt if not, source of growth matters

• Country fixed effects protect against bias from unobserved country heterogeneity within country estimates

• But heterogeneity in poverty reducing effects still possible (inequality and importance of extractive industry)

Estimating the participation effect Estimating the participation effect -- – empirical methodology – empirical methodology (2) (2)

•Augment previous equation with interaction terms:

With Xit-1=Gini coeff (GN), share in GDP of extractive industry (M), and initial income per capita (z/Y)

Similar test of equality (for different values of X variables)

If equality not rejected, then the equation collapses to a simple regression of the rate of poverty reduction on the rate of growth of GDP and source of growth would not matter.

Model estimated using World Bank PovCal data base for poverty episodes, and World Development Indicators for sectoral growth rates

Altogether 265 poverty spells in sample, drawn from 80 countries across the different continents over 1980-2002; 70 % of countries with more than one spell

About 1/3 of the poverty spells from 1980s; and 2/3 from 1990s (globalization and liberalization)

OLS with country fixed effects

Four poverty measures: $1-day poverty gap squared; $1-day poverty HC; $2-day poverty gap squared; $2-day HC

Estimating the participation effect Estimating the participation effect --

- - datadata

On average, growth from agriculture On average, growth from agriculture more poverty reducingmore poverty reducing

$1-day $2-dayCountry fixed effects estimates HC PG2 HC PG2

Coeff/p-value

Coeff/p-value

Coeff/p-value

Coeff/p-value

Agric growth/cap -7.08*** -5.9** -4.47*** -4.01***0.00 0.03 0.00 0.00

Agric growth/cap*Gini t-1 - - - -

Non-Agric growth/cap -1.78*** -2.10* -3.33*** -1.87***0.01 0.07 0.00 0.00

Non-Agric growth/cap*Gini t-1 - - - -

Non-Agric growth/cap*10% mining share - - - -

Agriculture more poverty reducing $1-Agriculture more poverty reducing $1-day; Nonagriculture more poverty day; Nonagriculture more poverty reducing $2-dayreducing $2-day

$1-day poverty $1-day $2-dayCountry fixed effects estimates HC PG2 HC PG2

Coeff/p-value

Coeff/p-value

Coeff/p-value

Coeff/p-value

Agric growth/cap -24.8*** -25.3*** -6.38 -9.76**0.00 0.01 0.12 0.04

Agric growth/cap*Gini t-1 50.92*** 49.62** 19.32** 21.25*0.00 0.02 0.04 0.05

Non-Agric growth/cap -4.63 1.67 -17.1*** -5.56**0.11 0.74 0.00 0.03

Non-Agric growth/cap*Gini t-1 7.76 -8.39 32.65*** 7.850.23 0.47 0.00 0.18

Non-Agric growth/cap*10% mining share 2.42 0.87 5.15*** 3.27**0.16 0.78 0.00 0.04

Participating in growth – differences Participating in growth – differences across sectors and degrees of across sectors and degrees of povertypoverty

Agriculture more powerful in reducing Agriculture more powerful in reducing $1-day poverty, with its advantage $1-day poverty, with its advantage declining as inequality increasesdeclining as inequality increases

Non-agriculture has the edge when it Non-agriculture has the edge when it comes to $2-day poverty, but not if comes to $2-day poverty, but not if driven by extractive industries and driven by extractive industries and also less effective in poorer countriesalso less effective in poorer countries

The Role of Agriculture in Poverty Reduction – towards a synthesis

pit aitsait yait(ynait-k) + nitsnit ynit(yait-k)

yait

ynait

Pit

aitsait

nits nit

Growth components

Indirect effectyait(ynait-k) &ynait(yait-k)

Participation and share components

Direct effect

Direct effect

Overall poverty reducing effect from Overall poverty reducing effect from growth across sectors – direct and growth across sectors – direct and indirect effectsindirect effects

Direct effect: Direct effect:

effects differ depending on level of Gini and presence of extractive effects differ depending on level of Gini and presence of extractive industryindustry

Indirect effect:Indirect effect:

Settings – stage of development: linkage effects evolve and so do Settings – stage of development: linkage effects evolve and so do sectoral shares: middle income, low income (excl SSA), low income sectoral shares: middle income, low income (excl SSA), low income SSA SSA with and w/o extractive industries with and w/o extractive industries

effects differ for middle & low income countries & SSA effects differ for middle & low income countries & SSA

Alternative definitions of sectoral Alternative definitions of sectoral growthgrowth

Comparing effects of one percent sectoral growth (as Comparing effects of one percent sectoral growth (as above) above) penalizes agriculture penalizes agriculture– AG largest individual sector, but smaller than rest of economy AG largest individual sector, but smaller than rest of economy

much less likely 1% ag GDP growth yielding similar amount much less likely 1% ag GDP growth yielding similar amount of OVERALL poverty reduction as 1% nag GDP growthof OVERALL poverty reduction as 1% nag GDP growth

resources needed to achieve 1% ag GDP growth hike likely resources needed to achieve 1% ag GDP growth hike likely much less than resources needed to achieve 1% nag GDP much less than resources needed to achieve 1% nag GDP growth; growth;

Lower bound!Lower bound!

effects differ depending on level of Gini and presence effects differ depending on level of Gini and presence of extractive industryof extractive industry

Alternative definitions of sectoral Alternative definitions of sectoral growth (2)growth (2)

Alternative Alternative control for size of sector and control for size of sector and explore effect of one percent aggregate GDP explore effect of one percent aggregate GDP growth coming from a sector (as done in the growth coming from a sector (as done in the literature)literature)

Implicitly assumes that it is as easy for a small Implicitly assumes that it is as easy for a small sector to generate 1% aggregate GDP growth sector to generate 1% aggregate GDP growth than for a larger sector than for a larger sector upper bound upper bound

Four key insights from Four key insights from combined effectscombined effects

1.1. Irrespective of setting, AG > 5 times more powerful in reducing Irrespective of setting, AG > 5 times more powerful in reducing poverty among poorest of the poor, as long as inequality not too highpoverty among poorest of the poor, as long as inequality not too high

2.2. Non-agriculture more powerful in reducing $2-day poverty HC if Non-agriculture more powerful in reducing $2-day poverty HC if extractive industry <10% of GDPextractive industry <10% of GDP

3.3. Agriculture 0.45 times (resource poor middle income) to 3.4 times as Agriculture 0.45 times (resource poor middle income) to 3.4 times as powerful in reducing $1-day poverty depending on sectoral or powerful in reducing $1-day poverty depending on sectoral or aggregate growth simulations; in SSA going from 1.7 to 3 times as aggregate growth simulations; in SSA going from 1.7 to 3 times as powerful (Gini< 0.45)powerful (Gini< 0.45)

4.4. Extractive industry substantially reduces poverty reducing power Extractive industry substantially reduces poverty reducing power from nag growth and agriculture loses its edge at high levels of from nag growth and agriculture loses its edge at high levels of inequalityinequality

Concluding Remarks Concluding Remarks (1)(1)

1.1. Contributions of a sector to poverty reduction : Contributions of a sector to poverty reduction : – direct growth component – growth potentialdirect growth component – growth potential– Indirect growth component – interlinkage effects Indirect growth component – interlinkage effects – Participation by poor in growth in the sector – inequality/miningParticipation by poor in growth in the sector – inequality/mining– Size of the sector in the economy – stage of developmentSize of the sector in the economy – stage of development

2.2. Insights by componentInsights by component– Slower agricultural growth follows from Engel’s Law and is not to be Slower agricultural growth follows from Engel’s Law and is not to be

equated with inherently lower productivityequated with inherently lower productivity

– Large interlinkage effects from agriculture decline as countries Large interlinkage effects from agriculture decline as countries develop, but have not yet disappeared in SSAdevelop, but have not yet disappeared in SSA

– Much larger participation by the very poor in growth from agriculture, Much larger participation by the very poor in growth from agriculture, but not by the $2-day poor and not in fundamentally unequal but not by the $2-day poor and not in fundamentally unequal societiessocieties

Concluding Remarks Concluding Remarks (2)(2)

3.3. Simultaneous considerations of indirect growth, Simultaneous considerations of indirect growth, participation, and share components shows that growth: participation, and share components shows that growth:

– in agriculture at least five times more powerful in reducing poverty in agriculture at least five times more powerful in reducing poverty among the poorest of the poor ($1-day poverty gap squared) among the poorest of the poor ($1-day poverty gap squared)

– In non-agriculture more powerful among better off poor ($2-day In non-agriculture more powerful among better off poor ($2-day headcount)headcount)

– In agriculture up to 3.2 times more effective in low income In agriculture up to 3.2 times more effective in low income countries, when accounting for sector size, with the advantage countries, when accounting for sector size, with the advantage diminishing as countries become richer and inequality becomes diminishing as countries become richer and inequality becomes highhigh

– Poverty reducing effects from nonagricultural growth decline Poverty reducing effects from nonagricultural growth decline substantially in the presence of extractive industriessubstantially in the presence of extractive industries

Concluding Remarks Concluding Remarks (3)(3)

4.4. Implications for agricultural sector Implications for agricultural sector strategies in light of global food strategies in light of global food supply challengesupply challenge

– Most countries increased investment in Most countries increased investment in agriculture as trust in market mediated agriculture as trust in market mediated food security has erodedfood security has eroded

– Enhancing agricultural productivity indeed Enhancing agricultural productivity indeed a valid entry point also to enhance growth a valid entry point also to enhance growth and poverty reduction in most low income and poverty reduction in most low income countries, especially in SSA and in mineral countries, especially in SSA and in mineral rich countriesrich countries

– To maximize the poverty reducing effects, To maximize the poverty reducing effects, smallholder agriculture appears to have smallholder agriculture appears to have an edgean edge

Jean Paul Remanoby selected by his village peers in sourthern Madagascar to be the first paricipant in a new programme offering improved farming techniques and modern inputs; IRIN Photo by Tomas de Mul

Thank you!Thank you!

UNU-WIDER WP 36 – The (Evolving) Role UNU-WIDER WP 36 – The (Evolving) Role of Agriculture in Poverty Reduction – An of Agriculture in Poverty Reduction – An Empirical Perspective, Empirical Perspective, http://www.wider.unu.edu/http://www.wider.unu.edu/