s11205 017 1581 9

Download as pdf or txt
Download as pdf or txt
You are on page 1of 25

Soc Indic Res (2018) 136:595–619

https://doi.org/10.1007/s11205-017-1581-9

Linear and Non-linear Relationships Between Growth,


Inequality, and Poverty in a Panel of Latin America
and the Caribbean Countries: A New Evidence
of Pro-poor Growth

Khalid Zaman1 • Sadaf Shamsuddin2

Accepted: 16 February 2017 / Published online: 23 February 2017


 Springer Science+Business Media Dordrecht 2017

Abstract The relationships between growth, inequality, and poverty is widely discussed
area in the development economics, which fairly overcrowded by linear and non-linear
growth components, however, while developing an index for pro-poor growth, the non-
linearity portion of growth has been widely ignored that address in this study by using a
panel of 18 selected Latin America and the Caribbean countries from 1981 to 2012. The
study proposed a new measure of pro-poor growth index, called ‘Poverty Interdependence
Growth Index (PIGI)’, which further extended in order to satisfy the monotonicity criterion
of pro-poor growth and poverty reduction, called ‘Poverty Interdependence Equivalent
Growth Rate (PIEGR)’. The results show that the impact of per capita survey income and
income inequality on poverty measures are ‘linear’ in nature when controlling the non-
linear components of growth, however, if this assumption is relaxed, the study doesn’t
established either ‘U-shaped’ and/or ‘asymptotic’ relationship between the variables. The
non-poverty measures including educational expenditures, health expenditures and popu-
lation growth significantly increases F–G–T measures of poverty. The estimates of PIGI
and PIEGR reveal that out of 18 countries, there are 4 countries shows highly pro-poor
growth, 11 countries shows negative pro-poor growth index (i.e., immiserizing growth
scenarios, where a positive growth increases poverty), and the remaining 3 countries shows
pro-rich. The study illustrates that our new measure of pro-poor growth index fairly pro-
vides conclusive findings.

Keywords Pro-poor growth  Non-linear growth components  Income inequality  Panel


random effect  System panel GMM

Jel Classification C33  I32

& Khalid Zaman


Khalid_zaman786@yahoo.com
1
Department of Economics, University of Wah, Quaid Avenue, Wah Cantt, Pakistan
2
Center of Undergraduate Studies for Girls Olaysha, College of Business Administration, King Saud
University, Riyadh, Saudi Arabia

123
596 K. Zaman, S. Shamsuddin

1 Introduction

Pro-poor growth is one of the widely discussed areas in the development literature. The
seminal work of Kuznets (1955) considered one of the foundation works for this debate,
which later down focused by number of scholars and international agencies that coined the
word i.e., ‘broad-based growth’ and ‘pro-poor growth’ (see, Chenery et al. 1974; World
Bank 1990; ADB 1999; Ravallion 2004 etc.). Afterward, the debate has prolonged and
spread in an academic and research arena and till now it consider the most impulsive and
pragmatic contest among the development economists. World Bank (2011) segregated the
pro-poor growth measures into three broad categories, i.e., (1) the studies falls in the
‘aggregate measures’ of pro-poor growth including ‘Growth Incidence Curve (GIC)’
suggested by Ravallion and Chen (2003), and ‘decomposition of growth and inequality’ by
Datt and Ravallion (1992), (2) the study falls in ‘absolute measures’ of pro-poor growth
including ‘rate of pro-poor growth’ suggested by Ravallion and Chen (2003), and (3) the
studies falls in the ‘relative measures’ of pro-poor growth including ‘poverty bias of
growth’ by McCulloch and Baulch (2000), ‘pro-poor growth index’ by Kakwani and Pernia
(2000), ‘poverty equivalent growth rate’ by Kakwani and Son (2002), and ‘poverty growth
curve’ by Son (2003).

1.1 Discussion on Income and Non-Income Poverty

The debate on monetary/income and non-income (human) poverty, both widely pro-
nounced in poverty research agenda for inclusive growth. The advocates of income poverty
approach argued that poverty is based on measure of monetary/income that evaluate with it
certain threshold income levels, if income poverty surpasses the threshold level, we do not
consider a person to be a poor while reverse is reserve for poor. Though, the classical
definition of poverty served many purpose in materialistic world, however, it consider as
an indirect approach to measure poverty. The number of studies measured pro-poor growth
by monetary indicators i.e., income or consumption, for example, Ravallion and Chen
(2003) measured the rate of pro-poor growth by using absolute terms and confirmed that
economic growth supports the poor and reduce poverty, while Kakwani and Pernia (2000)
and McCulloch and Baulch (1999) measured it by relative concepts, i.e., growth propor-
tionally benefits the poor more than the non-poor. The poverty is multidimensional phe-
nomenon; therefore, the income alone does not truly translate the poverty reduction and
pro-poor growth reforms. The direct measure of poverty is non-monetary measures that
look directly at the bundle of goods and it measures what people are lacking for it. It
contains some basic necessities goods like food, shelter, clothing, basic education,
healthcare facilities, safe water, good sanitation, etc. Klasen (2008) introduced the number
of non-monetary indicators in pro-poor growth scenario by using a country case study of
Bolivia and utilized ‘Growth Incidence Curve (GIC)’ approach to country’s non-monetary
indicators including education, health, and nutrition. The findings come to the following
conclusion that income growth was relatively pro-poor and strongly oppose to the absolute
pro-poor growth, while in non-income dimensions, growth generally in favor of the poor in
absolute sense and shows some considerable improvement in the education, health, and
nutritional indicators at countrywide. Grosse et al. (2008) further extended the non-income
dimensions of pro-poor growth indicators for Bolivia by using both the GIC approach and
poverty equivalent growth rate, and evaluated different basket of goods contain education,
health, and nutrition during 1989–1998. The results generally supported the relative sense

123
Linear and Non-linear Relationships Between Growth,… 597

of pro-poor growth using both the monetary and non-monetary income approach, while it’s
weakly supported in absolute sense. The overall results conclude that poverty either based
on income approach and/or non-income approach shows considerable improvement in
different dimensions of poverty, which is generally favorable in the context of Bolivia.
Tran et al. (2015) identified the disparities between monetary and non-monetary poverty
dimensions by using the Vietnam’s panel data for 2007, 2008, and 2010. The results
conclude that the person who have a greater market accessibility to get more benefits from
monetary income as compared to the others, as initial phase of development largely
contributed to the higher rate of income poverty reduction by higher growth, however, it is
not a sufficient condition that the same person perform well in non-income poverty
dimensions, as the person required more additional efforts and time to perform well in non-
income dimensions. Zaman (2015) expanded the non-income poverty measures in 21
welfare indicators by using four household surveys, i.e., 2002, 2006, 2008, and 2011. The
results show the significant differences between both the poverty measures in terms of
aggregation and distributional patterns. The results supported largely relative pro-poor
growth in non-income dimensions and weakly supported the absolute pro-poor growth in
few non-income dimensions. Bader et al. (2016) further identified the significant differ-
ences between the monetary income and multidimensional poverty (MDP) measures by
using the Lao household data of 2007–2008. The results of the study show some interesting
finding in the country context, i.e., the study identified a large pool of household are
‘‘overlooked poor’’ which were not identified by monetary income, while it is identified in
MDP measures. The higher deprivation score is in education and nutrition identified the
‘‘overlooked poor’’ in MDP measures while monetary income failed to identify the poor,
which lead to clear opposite direction between the two poverty measures. The results
further provoked that the correlation between MDP and income poverty is limited, which
further confirmed that income poverty does not serve as a true proxy for MDP measures
and vice versa.

1.2 Relationships Between Growth, Inequality and Poverty

The relationship between poverty, inequality and development traced by the classical work
of Ahluwalia (1976), which presented the crucial facts regarding the three above stated
elements in a cross-country setting, and proclaimed that relative inequality increases along
with an increase in economic development, which declines at the later stages of devel-
opment. This relationship is widely exhibit longer in the poorest countries group. This
causal relationship is attributed with the educational attainment, flexible labor market, and
population reduction strategies that helpful to reduce relative inequality by economic
development across nations. Bourguignon (2004) interlinked the poverty, growth, and
inequality in a functional form of triangular mode and concluded that wealth redistribution
policies from rich to poor may tend to improve economic growth through the channel of
flexible credit market. The redistributive policies matter for reduction in general poverty by
increasing country’s income, which further improves the income transfer channel for pro-
equality growth. Heshmati (2004) examined the causal relationships between growth,
inequality, and poverty by using the longitudinal data set of countries with longer time
period and found the convergence in the country’s per capita income, while divergence is
presence in income inequality data set, which is attributed due to homogeneity in the panel
of advanced countries and heterogeneity among less developed countries. The results
further indicate the heterogeneity in the policy propositions for poverty reduction, as
income inequality considerably affect the process of poverty reduction, which further

123
598 K. Zaman, S. Shamsuddin

affected the country’s development process. In industrialized countries, the wage differ-
ential gap substantially increases by increasing trade liberalization policies, while in
developing countries; trade liberalization policies support the labor market to reduce the
wage inequality. Finally, the study comes to the conclusion with the global U-shaped
Kuznets curve. Ravallion (2001) discussed the data limitations across cross-country cor-
relation that hides the development phase of poor in developing countries. The results
enforced the need of distribution—corrected rate of growth in average income that helpful
to expedite the process of pro-poor growth and poverty reduction. The study suggested
micro analysis of growth and income redistribution that would support growth-oriented
policies across countries. Adams (2004) selected the data set of 60 developing countries
and confirmed that economic growth promotes pro-poor growth policies by reduction in the
poverty in the absence of income inequality, however, the study pointed out the economic
efficiency of growth variables, which previously used different variables in diverse eco-
nomic settings. The results come to the conclusion that per capita GDP is not a significant
variable that translated into pro-poor growth and poverty reduction, while changes in
survey income considered the significant predictor that have a considerable impact on
poverty reduction across countries. Dollar and Kraay (2002) considered the large panel of
92 countries with a 40 year time period and found that trade openness, economic stability,
fiscal instruments, and private property rights considerably increases the income of the
residents and does not discriminate rich and poor in a country, which exhibit that economic
policies may not trickle down to the poor directly, while its indirectly transferred benefits
of economic growth to the poor as received initially richer for the dominating side. The
growth-enhancing strategies would be beneficial to support the process of pro-poor growth
and poverty reduction across nations. Ravallion (2005) discussed the two stand points of
poverty—inequality trade off by analyzing the 70 developing and transition economies in
1990s, i.e., the first point is the significant trade-off between lower poverty with lower
relative inequality, and the second point is the trade -off between absolute inequality and
poverty, which is associated between rising inequality and falling poverty. The study, in
general, concluded that there is not a significant trade-off between absolute poverty and
relative inequality due to low correlation between income and changes in the relative
inequality. Therefore, it is advisable to see the absolute inequality with reference of
poverty reduction for policy formulation. Dollar et al. (2015) collected the larger data set
of 115 countries with 40 year data period and concluded that income inequality is less
sensitive to the growth, which implies that cross-country variation in changes in social
welfare is largely attributed to growth in average incomes.
The overall discussion confirmed the strong correlation between growth, poverty and
inequality that widely recognized the trickledown theory in favor of poor under macroe-
conomic factors. This debate is prolonged in cross-country analysis which supported lar-
gely pro-poor growth framework under judicious income distribution. The policy to
support the poor in equitable mode confined the idea of pro-equality growth arguments
across the globe.

1.3 Pro-poor Growth Scenario

This study proposed a new measure of pro-poor growth index, called ‘Poverty Interde-
pendence Growth Index (PIGI)’, which included both the linear and non-linear growth
components, which previous ignored in the existing pro-poor growth indices, for example,
McCulloch and Baulch (2000) proposed ‘poverty bias of growth (PBG)’, which mainly
focused on reducing inequality. The PBG index captured ‘pure growth effect’ and ‘pure

123
Linear and Non-linear Relationships Between Growth,… 599

inequality effect’ in the absence of one and another. This index has some limitations, as
higher the value of PBG does not necessarily imply greater poverty reduction because
changes in poverty depends upon pure growth effect. In addition, the PBG mainly driven
by ‘linear symmetric poverty decomposition’ as suggested by Kakwani (2000), while both
the studies completely ignored the non-linear adjustments of growth phases that has a
considerable impact on the pro-poor growth reforms. Kakwani and Pernia (2000) proposed
‘pro-poor growth index (PPGI)’ that falls under the strict definition of pro-poor growth
with relative approach. This index based on poverty elasticity of growth and inequality and
compute total poverty elasticity, than its relative to the growth elasticity to form PPGI.
This index suggested certain value judgment threshold levels for assessment i.e., whether
the growth phase was poor or anti-poor, as if the PPGI value greater than the unity, the
growth process considered pro-poor or else anti-poor. This index has some limitations, as it
does not satisfied the monotonicity criterion of poverty reduction, while it does not account
the non-linear symmetric decomposition of poverty to capture the later stages of economic
development towards poverty reduction. Kakwani and Son (2008) proposed another index
of pro-poor growth, called ‘poverty equivalent growth rate (PEGR)’ that captures the gains
and losses of growth via distribution channel. The PEGR satisfied monotonicity criterion of
pro-poor growth and poverty reduction. The PEGR is pro-poor (anti-poor), when the value
of PEGR is greater (lesser) than the value of actual growth rate between two surveys
period. This index although satisfied both necessary and sufficient condition of pro-poor
growth and poverty reduction, however, non-linearity symmetric poverty decomposition
and distribution channel in the later stages of economic development is overlooked again.
Similarly, the other poverty reduction curves including ‘Growth Incidence Curve’ and
‘Poverty Growth Curve’ both based on stochastic dominance curve while ignoring poverty
line and poverty measures.
The PIGI and PIEGR is based on linear (see, Kakwani and Pernia 2000; Kakwani and
Son 2008; Son and Kakwani 2008) and non-linear systematic poverty decomposition (see,
McKinley 2009; Zaman 2016) combined together to form a new and relative measures of
pro-poor growth index (authors’ self extracted), which satisfied monotonicity criterion of
pro-poor growth and poverty reduction at the later stages of economic development, while
it falls in the strict definition of pro-poor growth under relative domain of growth index.
This methodology is applied on 18 selected Latin America and the Caribbean countries by
using international poverty line of US $1.90 per day. The study first estimated the long-run
relationship between growth, inequality, poverty, and human poverty measures under the
premises of linear and non-linear growth components, by using panel least square
regression, panel random effect, and system panel GMM estimator in a panel of selected
Latin America and the Caribbean countries for the period of 1981–2012. Later on, the
study using the same concept and extended non-linear symmetric poverty decomposition in
the pro-poor growth index, and form a new, stable, and relative measure of pro-poor
growth index called ‘poverty interdependence growth index’ and ‘poverty interdependent
equivalent growth rate’, which applied on individual countries of 18 selected Latin
America and the Caribbean countries to assess whether the economic development at the
later stages has impeding any impact on poverty reduction. For this reason, the study
included non-linearity component of growth in an existing pro-poor growth indices.

123
600 K. Zaman, S. Shamsuddin

2 Data Source and Methodological Framework

The following variables has been used in order to examine the long-run relationship
between growth, inequality, poverty, and non-poverty measures, i.e., average monthly per
capita income expenditures, Gini index (a measure of inequality), Foster–Greer–Thorbecke
(F–G–T) indices including poverty headcount, poverty gap, and squared poverty gap; non-
poverty measures include children out of school, expenditures on primary and secondary
education, life expectancy at birth, infant mortality rate, and population growth in a panel
of 18 selected Latin America and the Caribbean countries, namely, Argentina-Urban (data
set from 1991 to 2013), Belize (1993–1999), Bolivia (1990–2013), Brazil (1981–2013),
Chile (1987–2013), Columbia-Urban (1988–1991), Costa Rica (1981–2013), Dominican
Republic (1986–2013), El Salvador (1991–2013), Guatemala (1986–2011), Jamaica
(1988–2004), Mexico (1989–2012), Nicaragua (1993–2009), Panama (1989–2013), Para-
guay (1990–2013), Peru (1997–2013), Uruguay-Urban (1992–2005), and Venezuela
(1981–2006). The unified international poverty line i.e., US $1.90 per day on the basis of
purchasing power parity (PPP)—2005, has been set out for the selected Latin America and
the Caribbean countries. The data for poverty, income, and income inequality is taken from
POVCALNET published by World Bank (2015a), while the data for educational expen-
ditures, health expenditures, and population growth is taken from World Development
Indicators published by World Bank (2015b). Table 1 shows the list of variables and their
descriptions for ready reference.
The study started with the systematic link developed by Bourguignon (2004) regarding
the poverty-inequality-growth triangle i.e.,
Change in Poverty  FðGrowth; Inequality; change in DistributionÞ ð1Þ
which further annoying compliance by McKinley (2009) and suggested the determinants of
poverty reduction specification under cross-sectional regression settings i.e.,

Poverty reduction  f ðGrowth; Inequality;Income per capita lagged;Gini coefficient lagged;


Interaction between per capita income and change in inequality)
ð2Þ

This study compliment the McKinley (2009) specification of determinants of poverty


reduction by including square of income inequality and square of per capita income instead
of lagged per capita income and lagged Gini coefficient to determine both the linear and
non-linear growth components in the poverty-growth model specifications i.e.,
lnðpÞi;t ¼ b0 þ b1 lnðyÞi;t þ b2 lnðgÞi;t þ b3 lnðyÞ2i;t þ b4 lnðy  gÞi;t þ b5 lnðgÞ2i;t þ b6 lnðcÞit
þ ei;t
ð3Þ
where ‘p’ indicates F–G–T poverty indices including poverty headcount (p0), poverty gap
(p1), and severity of poverty (p2); ‘y’ indicates mean per capita monthly income expen-
diture from survey, ‘g’ indicates Gini index, c indicates non-poverty measures (including
children out of school, primary and secondary education expenditures, life expectancy at
birth, infant mortality rate, and population growth), ‘ln’ indicates natural logarithm, ‘i’
indicates 18 selected Latin America and the Caribbean countries, ‘t’ indicates different
household integrated countrywide surveys in a range between 1981 and 2013, and e is the
white noise error term.

123
Linear and Non-linear Relationships Between Growth,… 601

Table 1 List of variables


Variables Symbol Measurement

Poverty headcount P0 %
Poverty gap P1 %
Square poverty gap P2 %
Gini coefficient G %
Income Y Average monthly per capita income expenditures on the basis of
2005 purchasing power parity
Children out of school COOS % Of primary school age
Expenditures on primary EPE % Of government expenditure on education
education
Expenditures on ESE % Of government expenditure on education
secondary education
Life expectancy at birth LEB Years
Infant mortality rate IMR Per 1000 live births
Population growth POPG Annual %

The coefficient b1 and b2 shows the linear component of growth and inequality, while
b3 –b5 shows the non-linear growth and inequality components in Eq. (3). If b1 \0 and
b2 [ 0, it implies that the impact of economic growth and income inequality on poverty
would be ‘negative linear’, while, if b1 [ 0 and b2 \0, it implies that the relationship
would be ‘positive linear’. On the other way around, If b3 [ 0, b4 \0, and b5 [ 0, while
linear terms kept constant, it implies that the relationship between economic growth and
poverty would be ‘U-shaped’, while reverse is true in case of ‘inverted U-shaped’ rela-
tionship between poverty and economic growth under the premises of rising income
inequality. The relationship would be either ‘positive asymptotic’ when b1 [ 0, b2 \0,
b3 [ 0, b4 \0 and b5 [ 0, while it would be ‘negative asymptotic’ for the case i.e., b1 \0,
b2 [ 0, b3 [ 0, b4 \0, and b5 [ 0. Finally, the relationship would be either ‘positive
inverted asymptotic’, if and only if, b1 [ 0, b2 \0, b3 \0, b4 [ 0, and b5 \0, while
reverse is true for ‘negative inverted asymptotic; i.e., b1 \0, b2 [ 0, b3 \0, b4 [ 0, and
b5 \0. In conclusion, Eq. (3) would also indicate whether the relationship between growth,
inequality and poverty holds ‘linear’ relationship, ‘U-shaped’ relationship, and/or
‘asymptotic’ relationship in a panel of countries. Figure 1 shows the plots of level data for
ready reference.
The study employed panel OLS regression that ignores the cross-country-time-invariant
characteristics, while after confirmation the appropriate model specification by Hausman
test, the study used panel random effect regression that incorporates the time invariant
characteristics between the selected countries. Equation (4) shows the panel random effect
regressions’ model specification i.e.,
lnðpÞi;t ¼ b0 þ b1 lnðyÞi;t þ b2 lnðgÞi;t þ b3 lnðyÞ2i;t þ b4 lnðy  gÞi;t þ b5 lnðgÞ2i;t þ b6 lnðcÞit
þ si;t þ ei;t
ð4Þ
where s absorb time variant characteristics among the panel of selected countries.

123
602 K. Zaman, S. Shamsuddin

Y
P0 P1 P2 Average monthly per capita income expenditures
Poverty Headcount Ratio Poverty Gap Squared Poverty Gap (on the basis of 2005 PPP)
60 25 16 800

50
20
12 600
40
15
30 8 400
10
20
4 200
5
10

0 0 0 0
i = 18 Countries i = 18 Countries i = 18 Countries i = 18 Countries
T = 1981-2013 T = 1981-2013 T = 1981-2013 T = 1981-2013
(Household surveys in between the range - 264 observations) (Household surveys in between the range - 264 observations) (Household surveys in between the range - 264 observations) (Household surveys in between the range - 264 observations)

G COOS EPE IMR


Gini Coefficient Children out of school Expenditure on primary Education Infant Mortality rate
(a measure of inequality) (% of primary school age) (as % of government expenditure on education) (per 1,000 live births)
70 40 70 100

60 80
60 30

50 60
50 20
40 40

40 10
30 20

30 0 20 0
i = 18 Countries i = 18 Countries i = 18 Countries i = 18 Countries
T = 1981-2013 T = 1981-2013 T = 1981-2013 T = 1981-2013
(Household surveys in between the range - 264 observations)

LEB POPG ESE


Life expectancy at birth, total Population growth Expenditure on secondary education
(years) (annual %) (as % of government expenditure)
85 4 50

80
3 40
75

70 2 30

65 1 20
60
0 10
55

50 -1 0
i = 18 Countries i = 18 Countries
i = 18 Countries
T = 1981-2013 T = 1981-2013
T = 1981-2013

Fig. 1 Plots of the level data. Source: World Bank (2015a, b)

The study further employed ‘system panel Generalized Method of Moments (GMM)
estimator’, which addresses both the serial correlation problem and possible endogeneity
among the regressors. The study adopted the Arellano and Bond (1991) of GMM estimator
that includes lagged dependent variable along with lagged explanatory variables in dif-
ference operator to eliminate country specific differences by appropriate instrumental list.
Equation (5) shows the Arellano–Bond model specification i.e.,
lnðpÞi;t ¼ b0 þ b1 lnðpÞi;t1 þ lnðyÞi;t þ b2 lnðgÞi;t þ b3 lnðyÞ2i;t þ b4 lnðy  gÞi;t
ð5Þ
þ b5 lnðgÞ2i;t þ b6 lnðcÞit þ zi;t þ ei;t

where ‘z’ indicates lagged explanatory variables as an instrumental list.


In addition, Arellano–Bond model specifications gives certain other diagnostic tests
including Sargan–Hansen J-statistic for validating the instrumental lists, while AR(1), and
AR(2) test used for detecting serial correlations at first difference estimator. After
obtaining the parameter estimates, the study proposed a new measure of pro-poor growth
index, called ‘Poverty intensive Growth Index (PiGI)’, which are based upon both linear
and non-linear growth components.
Pro-poor growth definition falls in two broad categories, i.e., general definition of pro-
poor growth and strict definition of pro-poor growth. In general, pro-poor growth defined
as growth that benefits/supports the poor in terms of poverty falls, while the proponents of
strict definition argue that pattern of growth and distribution of income matters where
poverty declines. Ravallion and Chen (2003) support the general definition of pro-poor
growth, while Kakwani and Pernia (2000), Son (2004) and Kakwani and Son (2008) etc.
supports the strict definition of pro-poor growth. Son (2004) and Ravallion and Chen
(2003) supports the first order dominance condition for poverty reduction where poverty
line and poverty measures does not matter, while Kakwani and Pernia (2000) and Kakwani

123
Linear and Non-linear Relationships Between Growth,… 603

and Son (2008) etc. created the rate of index for poverty reduction under the natio-
nal/global poverty line and F–G–T measures of poverty measure. The pro-poor growth
index proposed by Kakwani and Pernia (2000) falls in the monotonicity criterion, as it
includes both the growth component and the mechanism to share the income flow from rich
to the poors via direct and indirect linkages.
The ‘poverty interdependence growth index’ based on strict definition of pro-poor
growth, where economic growth, judicious income distribution, and ‘joint interdepen-
dence’ of growth and inequality matters for poverty reduction that trickle down to the poor
as compared to the non-poor. More specifically, the proposed index executes the ‘full
approach’ of pro-poor growth, as growth process judged by linear and non-linear growth
components. This index satisfied the monotonicity criterion of pro-poor growth reduction,
as it implies both the linear and non-linear part of economic growth and income distri-
bution that satisfied both the necessary and sufficient condition of poverty reduction. The
study further extended the poverty interdependence growth index into poverty interde-
pendence equivalent growth rate that captured the gains/losses of growth rate due to
inequality change. As Kakwani and Son (2008) argued that gains reflect pro-poor growth,
while losses imply pro-rich growth or anti-poor growth. This index supports the funda-
mentals of poverty equivalent growth rate under the premises of non-linear growth and
inequality components in it. The poverty interdependence growth index (PIGI) comprises
both the linear and non-linear growth-inequality components to judge whether the growth
process is pro-poor or anti-poor i.e.,
(i) Linear and non-linear growth components under joint dependence of growth and
inequality:
Dpa =pa Dpa =pa Dpa =pa
g¼ þ 2 2þ ðiÞ
Dy=y Dy =y Dðy  gÞ=y  g
(ii) Linear and non-linear inequality components under joint dependence of growth
and inequality:
Dpa =pa Dpa =pa Dpa =pa
f¼ þ 2 2þ ðiiÞ
Dg=g Dg =g Dðy  gÞ=y  g
(iii) Total poverty elasticity:
d ¼gþf ðiiiÞ
(iv) Poverty Interdependence Growth Index (PIGI):
d
u¼ ðivÞ
g
(v) Poverty Interdependence Equivalent Growth Rate (PIEGR):
c ¼ c  u ðvÞ

where g is the linear and non-linear growth elasticity under the presence of joint depen-
dence of growth and inequality, f is the linear and non-linear inequality elasticity under the
presence of joint dependence of growth and inequality, d is the total poverty elasticity, u is
the poverty intensive growth index, c is poverty intensive equivalent growth rate, a is F-G-
T measures of poverty i.e., poverty headcount (p0), poverty gap (p1), and squared poverty
gap (p2), and c is the actual growth rate between two time period.

123
604 K. Zaman, S. Shamsuddin

This index is relatively better and wide-ranging in terms of explaining changes in


poverty over time, which derived on the basis of four main elements i.e., (i) Sum of linear
and non-linear poverty elasticity of growth, while inequality includes as an interactive term
with the economic growth, (ii) Sum of linear and non-linear poverty elasticity of
inequality, while economic growth includes as an interactive term with the income
inequality, (iii) Total poverty elasticity, which is the sum of (i) and (ii) elements, (iv)
Poverty interdependence growth index is the formulation of (iii) relative to element (i), and
finally, (v) Poverty interdependence equivalent growth rate is the joint product of actual
growth rate and poverty interdependence growth index (which implies gains/losses of
economic growth). This index satisfied the monotonicity criterion of pro-poor growth
index under the non-linear growth components and presented the results in a more gen-
eralized form of poverty reduction.

3 Results

Table 2 shows the descriptive statistics and correlation matrix among the growth,
inequality poverty, and non-poverty measures in a panel of selected Latin America and the
Caribbean countries. The results show that Gini index, P0, P1, and P2 has a median values
of 51.225, 8.155, 3.315 and 2.045 respectively. The poverty gap has a highest peak of the
distribution, followed by squared poverty gap, headcount ratio, average monthly per capita
income, and Gini index. The average monthly per capita survey income expenditure has an
average value of 378.276 US$, having a standard deviation value of 127.690 US$ with
positively skewed distribution. The mean value of children out of school, expenditures on
primary education, infant mortality rate, life expectancy at birth, population growth, and
expenditures on secondary education is about 6.134% of primary school age, 41.012% of
education expenditures, 25.167 per 1000 live births, 71.823 years, 1.515%, and 29.610%
of education expenditures.
Table 2 Panel-B shows the estimates of correlation coefficient and found that all the
poverty measures including P0, P1, and P2 has a positive correlation with the income
inequality (i.e., r = 0.592, r = 0.567, and r = 0.519 respectively) while negative corre-
lation with the per capita average monthly income (r = -0.731, r = -0.633, and
r = -0.552 respectively). The result implies that higher economic growth lead to decrease
poverty measures which confirmed the necessary condition of poverty reduction, while
higher income inequality substantially increase poverty measures, which does not con-
firmed the sufficient condition of poverty reduction. The correlation results exhibit that the
impact of higher economic growth to poverty reduction is greater than the impact of higher
income inequality to increase poverty measures, which partially supported the theory of
trickledown hypothesis in a panel of selected countries. The results further confirm the
negative correlation between per capita income and income inequality that is the desirable
condition of pro-poor growth process in a panel of selected Latin America and the Car-
ibbean countries. The impact of children out of school, primary education expenditures,
infant mortality rate, and population growth is positive on F–G–T measures of poverty
indices, i.e., higher the stated non-poverty measures, higher will be the poverty incidence,
while the impact of life expectancy at birth and secondary education expenditures on F–G–
T measures of poverty indices is negative, which exhibit that higher the stated non-poverty
measures, lesser will be the poverty incidence in a panel of countries. These results has
been empirically estimated in the subsequent analysis of regression apparatus, where linear

123
Table 2 Descriptive statistics and correlation matrix
P0 P1 P2 Y G COOS EPE IMR LEB POPG ESE

Panel-A
Mean 9.762 4.311 2.788 378.276 51.240 6.134 41.012 25.167 71.823 1.515 29.610
Median 8.155 3.315 2.045 359.375 51.225 4.559 38.900 21.350 72.267 1.496 30.286
Maximum 50.590 24.760 15.300 720.400 63.300 34.585 67.152 85.600 80.895 3.842 45.118
Minimum 0.450 0.020 0.000 106.970 34.420 0.034 23.471 7.400 55.107 -0.064 10.105
Std. Dev. 7.353 3.824 2.761 127.690 5.134 5.704 9.983 14.114 4.352 0.631 8.132
Skewness 1.497 1.776 1.919 0.657 -0.132 1.967 0.864 1.610 -0.835 0.206 -0.465
Kurtosis 6.888 7.394 7.331 2.831 2.782 8.489 2.975 5.770 4.115 4.120 2.565
Panel-B: correlation matrix
P0 1.000 0.955 0.875 -0.731 0.592 0.162 0.186 0.677 -0.653 0.434 -0.277
P1 0.955 1.000 0.978 -0.633 0.567 0.184 0.126 0.579 -0.570 0.430 -0.283
Linear and Non-linear Relationships Between Growth,…

P2 0.875 0.978 1.000 -0.552 0.519 0.190 0.071 0.483 -0.486 0.411 -0.274
Y -0.731 -0.633 -0.552 1.000 -0.243 -0.222 -0.368 -0.533 0.586 -0.328 0.370
G 0.592 0.567 0.519 -0.243 1.000 -0.067 0.045 0.437 -0.395 0.388 -0.033
COOS 0.162 0.184 0.190 -0.222 -0.067 1.000 0.352 0.150 -0.182 0.201 -0.448
EPE 0.186 0.126 0.071 -0.368 0.045 0.352 1.000 0.208 -0.230 0.233 -0.530
IMR 0.677 0.579 0.483 -0.533 0.437 0.150 0.208 1.000 -0.926 0.353 -0.284
LEB -0.653 -0.570 -0.486 0.586 -0.395 -0.182 -0.230 -0.926 1.000 -0.276 0.262
POPG 0.434 0.430 0.411 -0.328 0.388 0.201 0.233 0.353 -0.276 1.000 -0.321
ESE -0.277 -0.283 -0.274 0.370 -0.033 -0.448 -0.530 -0.284 0.262 -0.321 1.000
P0, indicates poverty headcount; P1, indicates poverty gap; P2, indicates squared poverty gap; ‘Y’, indicates income; ‘G’, indicates Gini coefficient; COOS, indicates children
out of school; EPE, indicates expenditures on primary education; IMR, indicates infant mortality rate; LEB, indicates life expectancy at birth, POPG, indicates population
growth; ESE, indicates expenditures on secondary education
605

123
606 K. Zaman, S. Shamsuddin

and non-linear growth components included in hierarchal regression nodes to confirm the
‘linear’, U-shaped’, and ‘asymptotic’ relationship between growth, inequality, and poverty
in a region. Figure 2 shows the plots of differenced data of the candidate variables for
ready reference.
Table 3 shows the estimates of panel random effect, panel OLS regression, and system
panel GMM estimator under the absence of non-linear growth components. The results of
panel random effect regression reveal that per capita income has a negative relationship
with the all F–G–T measures of poverty i.e., P0, P1, and P2, while income inequality has a
positive relationship with all three F–G–T measures of poverty. In terms of elasticity, the
results of poverty elasticity of growth and poverty elasticity of income inequality shows the
more elastic relationship, as the coefficient value is greater than the value of unity. The
results further imply that income inequality has a greater impact in order to increase all the
three F–G–T measures of poverty as compared to the per capita income for decreasing
poverty measures in a panel of selected Latin America and the Caribbean countries. The
similar results has been found in panel OLS regression apparatus, where one per cent
increase in the per capita income significantly decreases the headcount ratio by -1.883%,
poverty gap by -1.938%, and squared poverty gap by -1.683%, while income inequality
considerably increases poverty headcount by 3.911, 4.599, and 5.473% respectively. The
important point is to be noted that the impact of per capita income on poverty reduction in
panel OLS regression apparatus is less than the estimates of panel random effect, while the
impact of income inequality on escalating poverty in panel OLS regression is far greater
than the estimates of panel random effect regression. The intensity to increase income
inequality in panel OLS regression estimates is obvious due to ignore the country specific
and time variant shocks in a panel of selected countries. The results of system panel GMM
estimator shows that average household survey income has a significant and negative
relationship with the poverty headcount, i.e., if there is 1% increase in average household
income, poverty declines around 0.777% points, however, the impact of income inequality

D(P0) D(P1) D(P2) D(Y)


20 10 8 200
10 5 4 100
0 0 0 0
-10 -5 -4 -100
-20 -10 -8 -200
-30 -15 -12 -300
i = 18 countries i = 18 countries i = 18 countries i = 18 countries
T = 1981-2013 T = 1981-2013 T = 1981-2013 T = 1981-2013

D(G) D(COOS) D(EPE) D(IMR)


20 10 12 5
5 8 0
10
0 4 -5
0 -5 0 -10
-10 -4 -15
-10
-15 -8 -20
-20 -20 -12 -25
i = 18 countries i = 18 countries i = 18 countries i = 18 countries
T = 1981-2013 T = 1981-2013 T = 1981-2013 T = 1981-2013

D(LEB) D(POPG) D(ESE)


8 .8 8
6 .4 4
4 0
.0
2 -4
0 -.4 -8
-2 -.8 -12
i = 18 countries i = 18 countries i = 18 countries
T = 1981-2013 T = 1981-2013 T = 1981-2013

Fig. 2 Plots of differenced data. Note: ‘D’ indicates first difference. Source: World Bank (2015a, b)

123
Table 3 Results of panel random effect, Panel OLS regression, and panel GMM estimator for linear relationships between poverty, economic growth, income inequality, and
non-poverty measures
Variables Panel fixeda/Random effect regression Panel OLS effect System GMM (two step)
a a
ln(p0) ln(p1) ln(p2) ln(p0) ln(p1) ln(p2) ln(p0) ln(p1) ln(p2)

Constant -16.303* -36.159* -58.545* -3.110 4.011 17.278** – – –


ln(p0)t-1 – – – – – – 0.390* – –
ln(p1)t-1 – – – – – – 0.441* –
ln(p2)t-1 – – – – – – – – 0.451*
ln(y)t -2.173* -2.785* -1.904* -2.009* -2.229* -1.890* -0.777** -0.678 -0.138
ln(g)t 3.562* 4.550* 5.943* 3.449* 4.068* 4.637* 0.552 0.249 0.555
ln(COOS)t -0.008 0.042 0.113* -0.062* -0.079** -0.041 -0.023 0.012 0.038
ln(EPE)t 0.273 0.696*** 1.092* -0.160 -0.211 -0.772* 1.323* 1.802** 2.338*
ln(ESE)t 0.130 0.036 0.479 0.090 0.175 -0.144 -0.145 -0.091 0.208
ln(IMR)t 0.060 -0.088 0.528** 0.019 -0.415** -0.829* 0.423 0.248 0.508
Linear and Non-linear Relationships Between Growth,…

ln(LEB)t 3.587* 7.752* 9.215* 0.837 -1.027 -4.230* 2.834 1.919 1.931
ln(POPG)t 0.133* 0.299* 0.413* 0.192* 0.365* 0.575* 0.073 0.209 0.299***
Statistical tests
R-squared 0.862 0.844 0.902 0.876 0.723 0.709 – – –
Adjusted R-squared 0.858 0.827 0.891 0.872 0.714 0.700 – – –
F-statistics 198.440* 51.190* 86.418* 224.157* 82.722* 76.579 – – –
Hausman Test (Prob. value) 12.030 (0.149) 16.934 (0.030) 21.662 (0.005) – – – – – –
J-statistic – – – – – – 7.41E-26 4.72E-27 1.44E-26
Instrument rank – – – – – – 9 9 9
AR(1)-Prob. value – – – – – – 0.743 0.797 0.720
AR(2)-Prob. value – – – – – – 0.678 0.558 0.684
*, ** and *** indicates 1, 5 and 10% level of significance respectively. For system GMM Row (2 step) procedure, lagged dependent variable and lagged explanatory variables
are used as an instrumental variables
a
607

123
Shows panel fixed effect regression estimates
608 K. Zaman, S. Shamsuddin

on F–G–T measures of poverty incidence is statistically insignificant during the study time
period. The impact of children out of school is positive on squared poverty gap in panel
fixed effect regression while it has a negative impact on poverty headcount and poverty gap
in panel OLS settings. The impact of primary education, life expectancy at birth, and
population growth is positive in F–G–T measures of poverty indices, while the infant
mortality rate significantly influenced the squared poverty gap in a panel of selected
countries. The overall results imply that the impact of economic growth and income
inequality on poverty estimates confirmed the ‘negative linear’ relationship between them
under the absence of non-linear growth components. Ravallion (2001) argued that the
relationship between growth, inequality, and poverty in cross-country correlations required
more in depth micro analysis to confined their impact on each other, while available
evidence provoke that poors received greater benefits from rapid economic growth while
received lesser benefits from economic contraction due to unjustified economic distribu-
tion. Dollar and Kraay (2002) confirmed that aggregate affluence has a considerable impact
on poorest quintile for raising their average incomes that would helpful to reduce poverty
across the countries. De Dominicis et al. (2008) concluded the Meta analysis for growth—
inequality relationship and confirmed the differences of results due to diverse econometric
applications, quality of data, and sample selection that produced controversial results
between the two variables. White and Anderson (2001) argued that in majority of the cases
across the world regions, growth effect dominates, while in fewer cases changes in income
distribution affects poors income. The study concluded that although growth effect dom-
inates, however, it is necessary to hold pattern of income distribution judiciously for pro-
poor growth reforms across the countries. Yao (1999) discussed different findings for pro-
poor growth reforms in China, while the major conclusion is that poverty incidence sen-
sitized by growth and inequality component that hinders the flow of economic benefits
towards poor peoples.
Table 4 shows the estimates of non-linear growth components in growth-inequality-
poverty triangle by panel random effect, panel OLS regression and panel GMM estimator.
The results of panel random effect reveal that square of per capita income considerably
decreases F–G–T measures of poverty indices, while square of income inequality have a
positive relationship with all the three F–G–T poverty measures. The similar results has
been obtained by panel OLS regression, where square of per capita income significantly
reduces poverty headcount, while income inequality increases poverty in the latter stages
of economic development. The interaction term of growth and inequality are also used to
assess the non-linear relationship between the variables that confirmed the negative rela-
tionship with F–G–T measures of poverty indices. The results of system panel GMM
estimates show that square of per capita income significantly reduces poverty headcount
and squared poverty gap, while the square of income inequality has a positive relationship
with all three F–G–T measures of poverty. The overall results do not confine any rea-
sonable U-shaped relationship among the variables. The impact of non-poverty measures
including children out of school, primary and secondary education expenditures, and infant
mortality rate has a positive impact on squared poverty gap in a panel random effect
regression, while it further followed the positive relationships of life expectancy at birth
and population growth with the F–G–T measures of poverty across nations. Janvry and
Sadoulet (2000) confirmed the strong linkages between growth–inequality–poverty triangle
and concluded that economic growth impacts positively on poverty reduction and nega-
tively on increasing income inequality, however, under the premises of higher secondary
school education, economic growth have a considerable greater impact on urban poverty
reduction. Shahbaz (2010) concluded that economic growth has not a significant impact on

123
Table 4 Results of panel random effect, panel OLS regression, and panel GMM estimator for non-linear relationships between poverty, economic growth, income inequality,
and non-poverty measures
Variables Panel fixed effecta/random effect regression Panel OLS effect System GMM (two step)

ln(p0) ln(p1) ln(p2)a ln(p0) ln(p1) ln(p2) ln(p0) ln(p1) ln(p2)

Constant -18.959* -31.497* -57.235* -9.870** -6.795 11.427 – – –


ln(p0)t-1 – – – – – – -0.145 – –
ln(p1)t-1 – – – – – – – -0.163** –
ln(p2)t-1 – – – – – – – – -0.153
ln(y^2)t -0.709* -0.512** -1.257* -0.440* -0.212 -0.477 -0.821*** 0.121 -1.907**
(y 9 g)t -4.35E-05* -9.63E-05* 3.74E-05 -6.42E-05* -0.00010*
-5.40E-05*** -1.70E-05 -0.00016** 0.0001
ln(g^2) 2.062* 2.930* 2.736* 2.183* 2.768* 2.702* 1.403* 2.305* 1.937**
ln(COOS)t -0.009 0.023 0.115* -0.062* -0.079** -0.042 -0.158 0.062 -0.187
Linear and Non-linear Relationships Between Growth,…

ln(EPE)t 0.134 0.035 1.256* -0.247** -0.350** -0.832* -2.961 -4.125 -7.443
ln(ESE)t 0.095 0.153 0.518*** 0.056 0.121 -0.162 2.516 2.061 4.516
ln(IMR)t 0.027 -0.273 0.583** -0.040 -0.512* -0.879* -0.611 -1.618 1.368
ln(LEB)t 3.020** 4.061*** 9.802* 0.456 -1.637 -4.509* -5.735 -5.635 10.975
ln(POPG)t 0.136* 0.316* 0.407* 0.195* 0.370* 0.576* 0.307 0.640 1.010
Statistical tests
R-squared 0.868 0.719 0.903 0.884 0.737 0.712 – – –
Adjusted R-squared 0.863 0.709 0.892 0.880 0.728 0.702 – – –
F-statistics 184.672* 71.836* 83.748* 214.179* 78.633* 68.926* – – –
Hausman test (Prob. value) 9.400 (0.401) 12.320 (0.195) 22.777 (0.006) – – – – – –
J-statistic – – – – – – 3.43E-28 2.11E-28 1.34E-27
Instrument rank – – – – – – 10 10 10
AR(1)-prob. value – – – – – – 0.048** 0.192 0.573
609

123
Table 4 continued
610

Variables Panel fixed effecta/random effect regression Panel OLS effect System GMM (two step)
a
ln(p0) ln(p1) ln(p2) ln(p0) ln(p1) ln(p2) ln(p0) ln(p1) ln(p2)

123
AR(2)-prob. value – – – – – – 0.720 0.824 0.355

*, ** and *** indicates 1, 5, and 10% level of significance. For system GMM Row (2 step) procedure, lagged dependent variable and lagged explanatory variables are used as
an instrumental variables
a
Shows panel fixed effect regression estimates
K. Zaman, S. Shamsuddin
Linear and Non-linear Relationships Between Growth,… 611

inequality under the linear growth setting, while the study confirmed the existence of
Kuznets inverted U-shaped and inverted S-shaped relationship between growth and
inequality. Heshmati (2004) confirmed the global U-shaped Kuznets curve between eco-
nomic growth and income inequality, while the study further confirm the negative rela-
tionship between economic growth and income inequality across the countries.
Table 5 shows the estimates of linear and non-linear growth components in a single
regression apparatus and evaluated by panel random effect regression, panel OLS
regression, and system panel GMM estimator. The results of panel random effect reveal
that per capita income significantly reduces poverty head count and squared poverty gap,
while income inequality increases poverty headcount and poverty gap. Per capita income at
the later stages of development significantly reduces the poverty headcount, while inter-
action term decreases poverty gap in a region. In another regression apparatus, panel OLS
regression reveals that per capita income decreases the poverty headcount ratio, which
further prolonged to the later stages of economic development, as square of per capita
income significantly reduces the poverty headcount across nations. There is a monotonic
increasing relationship between income inequality and F–G–T measures of poverty indi-
ces, as income inequality increases poverty measures while the square of income inequality
does not show a significant relationship with the poverty indices. The estimates of panel
GMM confirm that the impact of household income expenditures and square of income on
poverty headcount and poverty gap is positive, while the interaction terms of income and
inequality reduces the F–G–T measures of poverty indices in a panel of countries. There is
a positive impact of children out of school, primary and secondary education expenditures,
and infant mortality rate on squared poverty gap, while the life expectancy and population
growth significantly influenced the F–G–T measures of poverty indices. Khan et al. (2016)
discussed the issues of labour market in terms of educational inequalities, health
inequalities and labor force unemployment, and emphasized the need of pro-equality and
pro-growth distribution for robust policy interventions to clear the market distortions
across nations. Ravallion and Chen (1997) concluded that economic growth supports to
reduce poverty in a good times while during recession, changes has been visible in poverty
incidence due to larger income inequality. The income distribution channel should be
checked and monitored during the good and bad times. Lin et al. (2009) argued that the
potential factor to hinders low income countries’ growth are the changes in the pattern of
flow of incomes between rich and poors, while higher inequality accelerate high -income
countries’ growth. The policies should be formulated to sustain broad based growth for
lower income countries, while policies to develop the income distribution channel in favor
of poors more than non-poors for high income countries.
The other diagnostic tests for the given model confirm that the Sargan-Hansen
J-statistics value is insignificant that favor the prescribed instrumental lists in the study. In
addition, serial correlation tests detected by AR(1) and AR(2) confirm that the error term at
the first difference are not correlated with the exogenous variables, therefore, we may
safely conclude that there is no problem of serial correlation in a given model. The
diagnostic tests for panel random effect and panel OLS regression indicate the goodness of
fit of the model by desired range of adjusted R-squared, along with significance F-statistics,
which confirm the model stability at 1% level of significance.
Table 6 shows the estimates of poverty interdependence pro-poor growth index and
poverty interdependence equivalent growth rate, which combines both the linear and non-
linear growth components in an existing pro-poor growth index. This index would be better
and comprehend from previous available pro-poor growth indices as its included non-linear
growth components to overcome the problem of non-linearity in pro-poor growth process.

123
Table 5 Results of panel random effect, Panel OLS regression, and panel GMM estimator for linear and non-linear relationships between poverty, economic growth, income
612

inequality, and non-poverty measures


Variables Panel fixed effecta/random effect regression Panel OLS effect System GMM (two step)

123
a
ln(p0) ln(p1) ln(p2) ln(p0) ln(p1) ln(p2) ln(p0) ln(p1) ln(p2)

Constant -8.450 -30.713** -44.298* -3.962 -7.446 10.471 – – –


ln(p0)t-1 – – – – – – 0.474* – –
ln(p1)t-1 – – – – – – – 0.345* –
ln(p2)t-1 – – – – – – – – 0.361*
ln(y)t -1.979* -0.759 -2.736* -1.423* -0.129 -0.914 6.731* 7.275*** 7.459
ln(g)t 2.953** 4.484*** 2.916 4.001* 5.021*** 5.575*** -4.680 -19.389 -29.829
(y^2)t -3.41E-06* 1.41E-06 1.79E-06 -2.81E-06* 1.29E-06 2.43E-07 6.59E-06* 9.76E-06*** 9.21E-06
(y 9 g)t 4.31E-05 -0.00013* 8.05E-05 1.38E-05 -0.00014* -6.04E-05 -0.00052* -0.00060** -0.00057***
(g^2)t -4.61E-05 0.0004 0.0003 -0.0002 0.0002 -1.07E-05 0.002 0.006 0.008
ln(COOS)t -0.010 0.026 0.114* -0.066* -0.078** -0.042 -0.063** -0.039 -0.007
ln(EPE)t -0.009 0.122 1.139* -0.253* -0.345*** -0.831* -0.343* 0.823 1.532
ln(ESE)t 0.100 0.120 0.535*** 0.048 0.112 -0.158 -0.248 -0.621 -0.443
ln(IMR)t -0.022 -0.274 0.515** -0.034 -0.521* -0.879* -0.291 -0.478 -0.453
ln(LEB)t 2.331*** 4.590** 9.193* 0.054 -1.436 -4.476* -0.793 0.666 0.975
ln(POPG)t 0.180* 0.295* 0.435* 0.187* 0.372* 0.577* 0.133* 0.283 0.453
Statistical tests
R-squared 0.875 0.720 0.904 0.888 0.738 0.712 – – –
Adjusted R-squared 0.870 0.708 0.892 0.883 0.726 0.700 – – –
F-statistics 160.360* 58.625* 78.069* 181.762* 64.074* 55.949* – – –
Hausman test (Prob. value) 7.191 (0.783) 11.324 (0.416) 18.563 (0.069) – – – – – –
J-statistic – – – – – – 8.79E-23 3.31E-23 1.68E-22
Instrument Rank – – – – – – 12 12 12
AR(1)-Prob. value – – – – – – 0.368 0.737 0.564
K. Zaman, S. Shamsuddin
Table 5 continued

Variables Panel fixed effecta/random effect regression Panel OLS effect System GMM (two step)
a
ln(p0) ln(p1) ln(p2) ln(p0) ln(p1) ln(p2) ln(p0) ln(p1) ln(p2)

AR(2)-Prob. value – – – – – – 0.566 0.640 0.595

*, ** and *** indicates 1, 5, and 10% level of significance respectively. For system GMM Row (2 step) procedure, lagged dependent variable and lagged explanatory variables
are used as an instrumental variables
a
Shows panel fixed effect regression estimates
Linear and Non-linear Relationships Between Growth,…
613

123
Table 6 Estimates of poverty interdependence growth index and poverty interdependence equivalent growth index
614

Countries Years AGR Poverty headcount (p0) Poverty gap (p1)

g f d u c g f d u c

123
Argentina-Urban 1991 and 2013 15.303 -9.658 8.008 -1.65 0.170 2.614 -8.66 7.944 -0.716 0.082 1.265
Belize 1993 and 1999 -21.260 -1.497 -2.955 -4.452 2.973 -63.228 -1.594 -1.453 -3.047 1.911 -40.641
Bolivia 1990 and 2013 82.284 -2.169 5.025 2.856 -1.316 -108.347 -0.136 9.145 9.009 -66.242 -5450.73
Brazil 1981 and 2013 80.888 -4.461 9.498 5.037 -1.129 -91.333 -3.939 7.705 3.766 -0.956 -77.336
Chile 1987 and 2013 96.552 -7.881 19.314 11.433 -1.450 -140.069 -7.881 19.314 11.433 -1.450 -140.069
Columbia-Urban 1988 and 1991 0.950 -6.032 2.054 -3.978 0.659 0.626 -8.779 0.38 -8.399 0.956 0.909
Costa Rica 1981 and 2013 15.126 -4.105 -7.921 -12.026 2.929 44.313 -4.249 -7.554 -11.803 2.777 42.017
Dominican Republic 1986 and 2013 8.048 -5.1 7.207 2.107 -0.413 -3.325 -5.37 8.796 3.426 -0.637 -5.135
El Salvador 1991 and 2013 38.248 -4.698 9.646 4.948 -1.053 -40.284 -5.816 15.194 9.378 -1.612 -61.673
Guatemala 1986 and 2011 309.976 -3.526 10.703 7.177 -2.035 -630.941 -4.194 13.747 9.553 -2.277 -706.056
Jamaica 1988 and 2004 96.647 -4.255 -8.417 -12.672 2.978 287.831 -4.218 -8.221 -12.439 2.949 285.017
Mexico 1989-2012 43.960 -6.357 7.678 1.321 -0.207 -9.135 -6.989 6.827 -0.162 0.023 1.018
Nicaragua 1993 and 2009 93.799 -3.957 5.017 1.06 -0.267 -25.126 -5.322 6.859 1.537 -0.288 -27.089
Panama 1989 and 2013 129.466 -7.799 15.967 8.168 -1.047 -135.592 -11.193 24.211 13.018 -1.163 -150.575
Paraguay 1989 and 2013 696.148 -4.354 10.859 6.505 -1.494 -1040.07 -3.703 13.381 9.678 -2.613 -1819.42
Peru 1997 and 2013 80.681 -7.962 7.089 -0.873 0.109 8.846 -9.931 9.178 -0.753 0.075 6.117
Uruguay 1992 and 2005 -346.081 -5.649 7.039 1.39 -0.246 85.1571 -3.315 2.905 -0.41 0.123 -42.803
Venezuela 1981 and 2006 -292.263 -7.082 -8.68 -15.762 2.225 -650.472 -14.062 -26.38 -40.442 2.875 -840.541
Countries Years AGR Squared poverty gap (p2)

g f d u c

Argentina-Urban 1991 and 2013 15.303 -7.642 7.62 -0.022 0.002 0.044
Belize 1993 and 1999 -21.260 -2.078 -0.434 -2.512 1.208 -25.701
Bolivia 1990 and 2013 82.284 1.52 12.655 14.175 9.325 767.354
K. Zaman, S. Shamsuddin
Table 6 continued

Countries Years AGR Squared poverty gap (p2)

g f d u c

Brazil 1981 and 2013 80.888 -3.323 6.218 2.895 -0.871 -70.470
Chile 1987 and 2013 96.552 -5.913 15.166 9.253 -1.56486 -151.091
Columbia-Urban 1988 and 1991 0.950 -10.984 -1.238 -12.222 1.112 1.057
Costa Rica 1981 and 2013 15.126 -4.19 -6.905 -11.095 2.6479 40.053
Dominican Republic 1986 and 2013 8.048 -5.146 10.078 4.932 -0.958 -7.714
El Salvador 1991 and 2013 38.248 -6.797 19.736 12.939 -1.903 -72.811
Guatemala 1986 and 2011 309.976 -4.485 14.88 10.395 -2.317 -718.439
Jamaica 1988 and 2004 96.647 -4.531 -8.59 -13.121 2.895 279.875
Mexico 1989–2012 43.960 -6.986 5.502 -1.484 0.212 9.338
Nicaragua 1993 and 2009 93.799 -6.221 8.523 2.302 -0.370 -34.709
Panama 1989 and 2013 129.466 -13.805 30.56 16.755 -1.213 -157.132
Linear and Non-linear Relationships Between Growth,…

Paraguay 1989 and 2013 696.148 -2.993 14.062 11.069 -3.698 -2574.56
Peru 1997 and 2013 80.681 -11.43 10.704 -0.726 0.063 5.124
Uruguay 1992 and 2005 -346.081 -0.809 -0.103 -0.912 1.127 -390.143
Venezuela 1981 and 2006 -292.263 -4.273 8.577 4.304 -1.007 -294.383

AGR indicates annual growth rate of mean dollar income between two household surveys. g, indicates poverty elasticity of growth; f, indicates poverty elasticity of inequality;
d, indicates total poverty elasticity; u, indicates poverty intensive growth index; c, indicates poverty intensive equivalent growth index; ‘p1’, indicates poverty gap; ‘p2’,
indicates squared poverty gap
615

123
616 K. Zaman, S. Shamsuddin

The results show that Argentina-Urban (Household survey data i.e., 1991 and 2013) is
intrinsically not considered as a pro-poor, although the country has a greater share of
poverty elasticity of growth as compared to the poverty elasticity of inequality, however,
the lower value of poverty intensive pro-poor growth index may not pronounce pro-poor
growth and poors not marginally benefited from existing growth reforms in a country.
Belize (1993 and 1999) although has a negative annual growth rate of -21.260, however,
this negative growth rate subsequently reduces income inequality that favor the poors. The
value of poverty intensive pro-poor growth index surpasses the threshold value of unity,
which indicates that the growth process was highly pro-poor and poors’ received marginal
benefits as compared to the non-poors in a country. The case study of Bolivia (1990 and
2013) is very interesting, as on one side, under the presence of poverty gap and squared
poverty gap, the larger share value of inequality and lower share value of economic growth
tends to become ‘immiserize’ the growth index, as the index value becomes negative,
while in case of severity of poverty, the positive and lower value of growth elasticity of
poverty, and positive and larger value of inequality elasticity of poverty surpasses the index
value is greater than unity, which exhibit that poors whom have a distance far from the
poverty line received marginal benefits as compared to the ultra poors and non-poors in a
country. The Columbia-urban (1988 and 1991), Costa Rica (1981 and 2013), and Jamaica
(1988 and 2004) are highly pro-poor countries, as the index value is greater than unity, and
poors received marginal gains from the existing growth reforms that were held in a par-
ticular countries. Venezuela (1981 and 2006) shows highly pro-poor growth index in case
of poverty head count and poverty gap, while the index value becomes less than the
threshold value of unity in case of squared poverty gap. One of the major reason for anti-
poor growth in case of severity of poverty is that the elasticity value of income inequality is
almost double than the elasticity value of per capita income that ‘immiserize’ the growth
index becomes negative. The remaining countries such as, Brazil (1981 and 2013), Chile
(1987 and 2013), Dominican Republic (1986 and 2013), El Salvador (1991 and 2013),
Guatemala (1986 and 2011), Mexico (1989 and 2012), Nicaragua (1993 and 2009),
Panama (1989 and 2013), Paraguay (1989 and 2013), and Uruguay (1992 and 2005), all
countries shows that the growth index is negative, which Bhagwati (1988) calls the ‘im-
miserizing’ growth phase. This growth phase is attributed because of two main reasons, at
first, high per capita income considerably increase F–G–T measures of poverty (rather than
reducing poverty), and secondly, the share of income inequality relative to increasing
poverty is far greater than the beneficial impact of per capita growth, which offsets the
growth effects relative to the rising inequality in a country profiles.
The overall results of poverty interdependence pro-poor growth index comparatively
more robust and pragmatic in a sense that it has included not only linear components of
growth, while it has included non-linear growth components, which previously largely
ignored while making pro-poor growth indices. The gains and/or losses of growth are
further assessed by the poverty interdependence equivalent growth rate, which merely
satisfied the monotonicity criterion of pro-poor growth and poverty reduction. Both the
indices fairly accompanied with the linear and non-linear growth components that previ-
ously crowded by the linearity proposition of pro-poor growth rates. The study provoke
that this new measure of pro-poor growth index provide better survey results as compared
to the available pro-poor growth indices in development literature.

123
Linear and Non-linear Relationships Between Growth,… 617

4 Conclusions

The objective of the study is to examine the linear and non-linear relationships between
growth, inequality, poverty, and non-poverty measures in a panel of 18 selected Latin
America and the Caribbean countries by utilizing the different household available surveys
for the periods of 1981–2012. The study proposed a new measure of pro-poor growth
index, called poverty interdependence growth index (PIGI), which are based on non-linear
symmetric poverty decomposition that evaluates the pro-poor growth reforms in the later
stages of economic development. This index further satisfied the monotonicity criterion of
pro-poor growth and poverty reduction, called ‘poverty interdependence equivalent growth
index (PIEGI)’ that evaluates the gains and/or losses of growth. The following key results
has been drawn by this exercise i.e.,
1. Panel correlation results confirm the ‘negative linear’ relationships between the
growth, inequality and poverty under the absence of non-linear growth components,
while this result further confirmed through panel fixed effect, panel least square
regression, and system panel GMM estimates.
2. While including both linear and non-linear growth components in growth-inequality-
poverty triangle, the study was not confined any reasonable form of relationships either
‘U-shaped’ or ‘asymptotic’ relationship between the variables.
3. The population growth, education, and health expenditures increase the F–G–T
measures of poverty indices across nations.
4. The study proposed a new form of pro-poor growth index, which includes both linear
and non-linear growth components in order to absorb the later stages of economic
development that were merely ignored in the available pro-poor growth indices.
5. This index, we called ‘poverty interdependence pro-poor growth index’ that falls in the
strict definition of pro-poor growth under relative approach, which merely based on
poverty line and poverty measures.
6. This index further extended to capture the gains and/or losses of growth that satisfied
both necessary and sufficient condition of poverty reduction. We judged this index by
the product (multiplication) of poverty interdependence pro-poor growth index and the
actual growth rate of survey income between two time periods, called poverty
interdependence pro-poor growth rate.
7. The results of proposed pro-poor growth indices show that out of 18 selected Latin
America and the Caribbean countries, only 4 countries show the highly pro-poor
growth, which confirm that the growth process facilitate the poors as compared to the
non-poors. We classified this situation as ‘trickle down’, where lower income strata
group get benefited from the existing growth reforms held in a countries.
8. Further, there are 11 countries that have a negative index value, which are merely two
main reasons i.e., firstly, the higher economic growth increases poverty, secondly, the
share of income inequality into increasing poverty is far greater than the impact of
economic growth on poverty, which we referred this situation as ‘immiserizing’
growth phenomenon in particular countries.
9. Finally, there are three countries that shows anti-poor growth, where growth process
benefited the non-poors as compared to the poors in their countries.
The study concludes that economic growth does not necessarily translate the poverty
reduction; if and only if, there should be judicious income distribution channels that
facilitate and expedite the process of pro-poor growth reforms across the countries. The

123
618 K. Zaman, S. Shamsuddin

policies should be formulated in order to sustained broad-based growth along with rational
income distribution channel that would get marginally benefited the poors as compared
with the non-poors. For broad-based growth, it is necessary to include social expenditures
in pro-poor growth agenda, which would facilitate the poors to escape out from poverty;
while for rational income distribution, the policy makers and government officials required
a comprehend taxation policies that should be flexible and elastic, and fulfill the basis of
pro-equality growth notions.

Acknowledgements The authors are thankful to the Deanship of Scientific Research, King Saud University,
Riyadh Saudi Arabia for funding research through research group project RG-1435-075.

References
Adams, R. H. (2004). Economic growth, inequality and poverty: Estimating the growth elasticity of poverty.
World Development, 32(12), 1989–2014.
ADB. (1999). Fighting poverty in Asia and the Pacific: The poverty reduction strategy. Manila: Asian
Development Bank.
Ahluwalia, M. S. (1976). Inequality, poverty and development. Journal of Development Economics, 3(4),
307–342.
Arellano, M., & Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an
application to employment equations. The Review of Economic Studies, 58(2), 277–297.
Bader, C., Bieri, S., Wiesmann, U., & Heinimann, A. (2016). Differences between monetary and multidi-
mensional poverty in the Lao PDR: Implications for targeting of poverty reduction policies and
interventions. Poverty & Public Policy, 8(2), 171–197.
Bhagwati, J. N. (1988). Poverty and public policy. World Development, 16(5), 539–654.
Bourguignon, F. (2004). The poverty-growth-inequality triangle. New Delhi: Indian Council for Research on
International Economic Relations.
Chenery, H., Ahluwalia, M. S., Duloy, J. H., Bell, C. L. G., & Jolly, R. (1974). Redistribution with growth;
policies to improve income distribution in developing countries in the context of economic growth.
Oxford: Oxford University Press.
Datt, G., & Ravallion, M. (1992). Growth and redistribution components of changes in poverty measures: A
decomposition with applications to Brazil and India in the 1980s. Journal of Development Economics,
38(2), 275–295.
De Dominicis, L., Florax, R. J., & De Groot, H. L. (2008). A meta-analysis on the relationship between
income inequality and economic growth. Scottish Journal of Political Economy, 55(5), 654–682.
Dollar, D., Kleineberg, T., & Kraay, A. (2015). Growth, inequality and social welfare: Cross-country
evidence. Economic Policy, 30(82), 335–377.
Dollar, D., & Kraay, A. (2002). Growth is good for the poor. Journal of Economic Growth, 7(3), 195–225.
Grosse, M., Harttgen, K., & Klasen, S. (2008). Measuring pro-poor growth in non-income dimensions.
World Development, 36(6), 1021–1047.
Heshmati, A. (2004). Growth, inequality and poverty relationships. IZA Discussion Paper No. 1338, Bonn,
Germany. https://papers.ssrn.com/sol3/papers2.cfm?abstract_id=602466. Accessed 12 Jan 2015.
Janvry, A. D., & Sadoulet, E. (2000). Growth, poverty, and inequality in Latin America: A causal analysis,
1970–94. Review of Income and Wealth, 46(3), 267–287.
Kakwani, N. (2000). On measuring growth and inequality components of poverty with application to
Thailand. Journal of Quantitative Economics, 16(1), 67–80.
Kakwani, N., & Pernia, E. M. (2000). What is pro-poor growth? Asian Development Review, 18(1), 1–16.
Kakwani, N., & Son, H. (2002). Pro-poor growth: Concept, measurement, and application. Sydney:
Unpublished mimeo, University of New South Wales.
Kakwani, N., & Son, H. H. (2008). Poverty equivalent growth rate. Review of Income and Wealth, 54(4),
643–655.
Khan, H. U. R., Khan, A., Zaman, K., Nabi, A. A., Hishan, S. S., & Islam, T. (2016). Gender discrimination
in education, health, and labour market: A voice for equality. Quality & Quantity. doi:10.1007/s11135-
016-0384-4.
Klasen, S. (2008). Economic growth and poverty reduction: Measurement issues using income and non-
income indicators. World Development, 36(3), 420–445.

123
Linear and Non-linear Relationships Between Growth,… 619

Kuznets, S. (1955). Economic growth and income inequality. The American economic review, 45(1), 1–28.
Lin, S. C., Huang, H. C., Kim, D. H., & Yeh, C. C. (2009). Nonlinearity between inequality and growth.
Studies in Nonlinear Dynamics & Econometrics, 13(2), 1–12.
McCulloch, N., & Baulch, B. (1999). Tracking pro-poor growth New ways to spot the biases and benefits.
Insights, 31(3), 1–10.
McCulloch, N., & Baulch, B. (2000). Simulating the impact of policy upon chronic and transitory poverty in
rural Pakistan. The Journal of Development Studies, 36(6), 100–130.
McKinley, T. (2009). Revisiting the dynamics of growth, inequality and poverty reduction. Centre for
Development Policy and Research Discussion Paper, 25(09). http://www.soas.ac.uk/cdpr/publications/
papers/file53130.pdf. Accessed 3rd Mar 2016.
Ravallion, M. (2001). Growth, inequality and poverty: Looking beyond averages. World Development,
29(11), 1803–1815.
Ravallion, M. (2004). Pro-poor growth: A primer. World Bank policy research working paper, (3242).
Ravallion, M. (2005). A poverty-inequality trade off? The Journal of Economic Inequality, 3(2), 169–181.
Ravallion, M., & Chen, S. (1997). What can new survey data tell us about recent changes in distribution and
poverty? The World Bank Economic Review, 11(2), 357–382.
Ravallion, M., & Chen, S. (2003). Measuring pro-poor growth. Economics Letters, 78(1), 93–99.
Shahbaz, M. (2010). Income inequality-economic growth and non-linearity: A case of Pakistan. Interna-
tional Journal of Social Economics, 37(8), 613–636.
Son, H. H. (2003). A new poverty decomposition. The Journal of Economic Inequality, 1(2), 181–187.
Son, H. H. (2004). A note on pro-poor growth. Economics Letters, 82(3), 307–314.
Son, H. H., & Kakwani, N. (2008). Global estimates of pro-poor growth. World Development, 36(6),
1048–1066.
Tran, V. Q., Alkire, S., & Klasen, S. (2015). Static and dynamic disparities between monetary and multi-
dimensional poverty measurement: Evidence from Vietnam. In I. G. Thesia & S. S. Kathleen (Eds.),
Measurement of poverty, deprivation, and economic mobility (Research on Economic Inequality) (Vol.
23, pp. 249–281). Bingley: Emerald Group Publishing Limited.
White, H., & Anderson, E. (2001). Growth versus distribution: Does the pattern of growth matter?
Development Policy Review, 19(3), 267–289.
World Bank. (1990). World development report 1990. New York: Oxford University Press.
World Bank. (2011). Measuring pro-poor growth. Washington, DC: PovertyNet, World Bank.
World Bank. (2015a). POVCALNET. Washington, DC: World Bank.
World Bank. (2015b). World development indicators. Washington, DC: World Bank.
Yao, S. (1999). Economic growth, income inequality and poverty in China under economic reforms. The
Journal of Development Studies, 35(6), 104–130.
Zaman, K. (2015). Measurement issues of income and non-income welfare indicators: Assessment of
Pakistan’s Pro-Poor growth. International Journal of Economics and Financial Issues, 5(3), 802–811.
Zaman, K. (2016). Examination of non-linear dynamics in Pro-poor growth framework: A case study of
Pakistan. International Journal of Economics and Empirical Research, 4(4), 207–211.

123

You might also like