The Effect of the Sectoral Composition of Economic Growth on Rural and Urban Poverty*

• Published date: 01 March 2021
• Author: Rui Benfica,Heath Henderson
• Date: 01 March 2021
• DOI: http://doi.org/10.1111/roiw.12462

© 2020 International Association for Research in Income and Wealth

248

THE EFFECT OF THE SECTORAL COMPOSITION OF ECONOMIC

GROWTH ON RURAL AND URBAN POVERTY*

by Rui benfica

International Food Policy Research Institute

AND

HeatH HendeRson†

Drake University

We examine the relationship between the sectoral composition of economic growth and the rural-urban

composition of poverty. To this end, we use a cross-country panel dataset consisting of 146 rural and

urban poverty “spells” for 70 low- and middle-income countries. We find that rural (urban) poverty is

highly responsive to agricultural (non-agricultural) productivity growth. The effect of agricultural pro-

ductivity growth on rural poverty is particularly strong for countries with little dependence on natural

resources. We also find that growth in the share of employment in the non-agricultural sector (i.e. struc-

tural transformation) reduces rural poverty, most notably for countries at a low initial level of develop-

ment. These findings are robust to changes in key assumptions, including using alternative poverty lines.

Finally, we use our estimates to examine the past contribution of different sources of economic growth

to rural and urban poverty reduction across regions.

JEL Codes: I32, O11, O47

Keywords: agriculture, economic growth, poverty, structural transformation

1. intRoduction

An understanding of the channels through which economic growth reduces

poverty is instrumental for promoting inclusive and sustainable economic develop-

ment. While it is well documented that growth tends to contribute to poverty

reduction, the empirical literature suggests that there is considerable heterogeneity

in the relationship across space and over time.1 For example, using data from the

1See Foster and Székely (2008), Ferreira et al. (2010), Ram (2011), or Chambers and Dhongde

(2011) for comprehensive reviews of the literature on the growth elasticity of poverty reduction.

*The authors would like to thank the editor and three anonymous referees for their comments. We

would also like to thank Shaohua Chen, Rebecca Ray, and Sean Severe for providing detailed com-

ments, and the International Fund for Agricultural Development (IFAD) for financial support. We are

additionally grateful for questions raised by attendees of the 43rd and 44th Annual Conference of the

Eastern Economic Association, Drake University’s “First Friday Brown Bag Seminar,” and partici-

pants in the December 2017 Workshop of the STAARS (Structural Transformation of African

Agriculture and Rural Spaces) project.

†Correspondence to: Heath Henderson, Drake University, 311 Aliber Hall, Des Moines, IA 50311,

USA (heath.henderson@drake.edu).

Review of Income and Wealth

Series 67, Number 1, March 2021

DOI: 10.1111/roiw.12462

bs_bs_banner

Review of Income and Wealth, Series 67, Number 1, March 2021

249

© 2020 International Association for Research in Income and Wealth

1980s to 1990s, Besley and Burgess (2003) showed that the growth elasticity of

poverty reduction varies across different regions, with elasticities ranging from

−0.49 (for sub-Saharan Africa) to −1.14 (for Eastern Europe and Central Asia). To

cite another example, Datt et al. (2016) examined changes in the growth elasticity

of poverty across time using data from pre- and post-reform India. Across a vari-

ety of specifications, they found that the responsiveness of poverty to economic

growth was significantly greater for the post-reform period.

Explanations for the observed heterogeneity have emphasized differences in

“initial conditions” and “patterns of growth.” Initial income inequality has fea-

tured prominently in analyses of initial conditions as higher inequality (1) may

slow the rate of growth (the “induced-growth argument”) and (2) may reduce sub-

sequent gains to the poor from existing growth (the “growth elasticity argument”)

(Ravallion, 1997). While evidence for the induced-growth argument is highly con-

text dependent (Neves and Silva, 2014), a number of studies have found support

for the growth elasticity argument (Kalwij and Verschoor, 2007; Ravallion and

Chen, 2007; Fosu, 2009). Bourguignon (2003) further showed that the growth elas-

ticity of poverty relates directly to the ratio of the poverty line to mean income.

That the responsiveness of poverty to growth increases with per capita income has

been corroborated by multiple subsequent empirical studies (Kalwij and Verschoor,

2007; Fosu, 2009; Chistiaensen et al., 2011).2

Regarding patterns of growth, Montalvo and Ravallion (2010) discuss two

reasons why the sectoral and/or geographic composition of economic activity

affects the growth-poverty relationship: (1) economic growth may occur in sectors

or locations that do not benefit poor people and (2) the composition of economic

activity can affect income inequality, which has implications for the subsequent

effect of growth on poverty (see above). There is a large body of empirical research

that finds that growth in the agricultural sector is particularly effective at reduc-

ing poverty, not only through its direct effect via agricultural incomes but also

indirectly through growth linkages with the rest of the economy (Bezemer and

Headey, 2008; Dercon, 2009; de Janvry and Sadoulet, 2010; Chistiaensen et al.,

2011). The responsiveness of poverty to agricultural growth, however, has been

found to diminish with development (Ravallion and Datt, 2002; Ferreira et al.,

2010; Chistiaensen et al., 2011).

The agriculture versus non-agriculture dimension has been a focus in the pat-

terns of growth literature, but other dimensions have been considered as well.

Using data from India, Ravallion and Datt (1996) decomposed mean consumption

growth into rural and urban components, and found that rural consumption

growth was the primary driver of poverty reduction.3 To the contrary, Datt et al.

2While the study of initial conditions has focused on initial inequality and the level of develop-

ment, a number of other factors have been explored. See Datt and Ravallion (1998) on infrastructure

and human resources, Ravallion and Datt (2002) on literacy and farm productivity (among other fac-

tors), Suryahadi et al. (2009) on human capital, Ferreira et al. (2010) on human development and

worker empowerment, and Chistiaensen et al. (2011) on the share of extractive industries in GDP.

3See Ravallion and Chen (2007) for a similar result in the context of China. This result is in part a

direct implication of the fact that baseline levels of poverty were much higher in rural areas where most

people live. Noticeable overall reductions in poverty could only be driven by improvements in consump-

tion in those areas and predominantly driven by agricultural growth.

Review of Income and Wealth, Series 67, Number 1, March 2021

250

© 2020 International Association for Research in Income and Wealth

(2016) found that urban growth came to occupy the leading role in the wake of

India’s reforms of the early 1990s. Suryahadi et al. (2009) took this geographical

decomposition a step further by decomposing rural and urban growth by economic

sector. Using data from Indonesia, they found that provincial poverty was particu-

larly responsive to growth in the urban and rural services sectors. Finally, in a

unique contribution, Loayza and Raddatz (2010) found, using cross-country data,

that the composition of growth in terms of the intensive use of unskilled labor is

critical for poverty reduction.4

While substantial progress has been made towards understanding the

growth-poverty relationship, the literature offers an incomplete characterization

of the channels through which growth reduces poverty. To what extent does overall

and sectoral growth have a differential effect on rural and urban poverty? Are these

growth effects driven by labor productivity growth or employment expansion? Are

employment expansion effects due to labor force growth or the movement of labor

across sectors (i.e. structural transformation)? How do initial conditions, partic-

ularly differences in economic inequality and the level of development, influence

the above channels? We examine these questions using a novel dataset consisting

of 146 rural and urban poverty “spells” for 70 low- and middle-income countries

spanning from 1992 to 2013. To the best of our knowledge, our dataset represents

the most comprehensive source of internationally-comparable rural and urban

poverty measures compiled to date.

Our primary contribution is the analysis of the relationship between the sec-

toral composition of growth and the rural-urban composition of poverty. Previous

research in this area has focused on single-country studies, largely in an Asian

context. Research on China has highlighted the association between agricultural

growth and rural poverty reduction (Ravallion and Chen, 2007; Montalvo and

Ravallion, 2010), whereas work on India and Indonesia has found that both rural

and urban poverty are responsive to growth in the agricultural and services sectors

(Ravallion and Datt, 1996; Suryahadi et al., 2009).5 In contrast to these studies, we

believe we are the first to address these questions using cross-country panel data.

Our dataset is also particularly rich in terms of covering a large number of coun-

tries over a relatively long period. While cross-country studies have well-known

limitations, we are able to provide new insights into the dynamics of rural and

urban poverty by examining other contexts (e.g. sub-Saharan Africa) and exploit-

ing cross-country variation in key variables (e.g. income inequality and level of

development).

In addition to providing complementary insights through the use of cross-coun-

try panel data, we seek to deepen the analysis of the relationship between sectoral

growth and the rural-urban composition of poverty in three ways. First, we exam-

ine the growth-poverty channels by decomposing sectoral growth into components

associated with labor productivity growth and employment expansion. Second, we

further decompose the employment expansion effects into components associated

4Again, this result is in part a direct effect derived from the fact that most of the poor are exactly

the unskilled that are available to work in a growing economy.

5Recent research suggests, however, that the sectoral composition of growth in India has become

less important for poverty reduction after the reforms of the early 1990s (Datt et al., 2016).