Escolar Documentos
Profissional Documentos
Cultura Documentos
by
A THESIS
in
ECONOMICS
MASTER OF ARTS
Approved
Accepted
August, 2012
Copyright 2010, Shahidur Rashid Talukdar
Texas Tech University, Shahidur Rashid Talukdar, August 2012
ACKNOWLEDGEMENTS
I am extremely grateful to Dr. Masha Rahnama for being the Chair of my thesis
committee and also for guiding me throughout the process of conducting empirical
analysis as well as writing the thesis. I would like to thank Dr. Victor Valcarcel for being in
my thesis committee and providing me with relevant ideas, helpful comments, and
suggestions. Last but not the least, I want to thank my wife, Mumtaz B. Laskar, for
helping me with processing the data, and providing me the emotional support to complete
my thesis. Without their help my thesis would not have been complete.
ii
Texas Tech University, Shahidur Rashid Talukdar, August 2012
TABLE OF CONTENTS
ACKNOWLEDGEMENTS .................................................................................................. ii
ABSTRACT.............................................................................................................. v
LIST OF TABLES................................................................................................................... vi
INTRODUCTION ............................................................................................................ 1
DATA.............................................................................................................................14
RESULTS ......................................................................................................................21
iii
Texas Tech University, Shahidur Rashid Talukdar, August 2012
REFERENCES ..............................................................................................................50
APPENDICES
iv
Texas Tech University, Shahidur Rashid Talukdar, August 2012
ABSTRACT
The aim of this thesis is to study the effect of inflation on poverty in developing
countries. I analyze the effect of inflation on poverty with a panel dataset comprised of
115 developing countries over the period 1981 - 2008. The dataset comprises 10
observations for each country as the data is available at 3 year intervals. As previous
studies indicate that poverty is also affected by factors such as income, external debt,
educational attainment, and quality of governance, besides inflation, I take these as
independent variables and poverty as the dependent variable. With the help of regression
analysis, I find evidence supporting the view that inflation in general is positively
correlated with poverty while income, educational attainment, and quality of governance
show negative correlation with poverty in most of the specifications. Apart from the study
of all the countries combined, I separately analyze the effect of inflation on poverty in low
income countries, lower middle income countries, and upper middle income countries to
see whether the effect of inflation is similar or different in countries with different levels of
income. I find that although in most of the cases inflation shows a positive and statistically
significant correlation with poverty, however, in the case of low income countries, the
relationship between inflation and poverty is negative and statistically insignificant under
certain specifications.
v
Texas Tech University, Shahidur Rashid Talukdar, August 2012
LIST OF TABLES
1. Summary Statistics 14
3. Descriptive Statistics 16
12.Summary Results 44
vi
Texas Tech University, Shahidur Rashid Talukdar, August 2012
CHAPTER 1
INTRODUCTION
1.1. MOTIVATION
When the prices of essential goods or services rise, it becomes harder for the
poor with a limited budget to be able to afford those goods and services. The limited
purchasing power of the poor shrinks when prices of essential commodities increase but
the income does not increase at the same pace (Wilson, 2011). To quote Wolfgang
Fengler, the World Banks lead economist in Kenya:
Imagine you are spending half of your income on something whose price suddenly
increases by a quarter. Seems impossible? This is how in fact inflation has hit the poor in
many developing countries, especially Kenya... So if you are exposed to high inflation,
there is no choice but to cut down on food or on other expenses, many of which are vital,
such as school fees or health care. This is why inflation is the worst tax on the poor.
(Blog post on Oct 03, 2011, Taxing the poor through inflation, Africa CanEnd Poverty A
World Bank Blog.)
If inflation hurts the poor in Kenya, what about the poor in other countries? Does
inflation hurt the poor everywhere else as well? Can inflation reduce people's
consumption because of higher prices of goods and services? Does that lead to an
increased level of poverty? Is the effect of inflation on poverty similar in all countries or is
it different? Questions such as these motivated me to carry out a study on the effect of
inflation on poverty.
1.2. BACKGROUND
Economists hold diverse opinions about the impact and consequences of inflation
on peoples well-being. Although it is generally accepted that extremely high inflation
rates, such as those during hyperinflation, affect the economy negatively, many
economists contend that low inflation is good for the economy. Howitt (1990), for
instance, argues that although inflation is costly, once it is under way, society is better off
to tolerate a little inflation than to bear the cost necessary to achieve price stability.
Summers (1991) and DeLong and Summers (1992) argue that moderate inflation is good
for the economy. Analyzing similar arguments of a number of economists, Marty and
Thornton (1995) concluded that a group of economists believe that moderate inflation is
good for the economy.
1
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Although there may be some economists who would argue that in the long run
inflation does not matter, however, in reality, inflation is a serious concern for common
people, i.e., non-economists. For example, Shiller (1996) found that there was a stark
difference between how common people and economists perceive inflation. Shiller found
that among non-economists, the largest concern with inflation appears to be that it lowers
peoples standard of living. Another study by Easterly and Fischer (2001) confirmed that
inflation concerns the poor more than the rich. Using polling data for 31,869 households
in 38 countries and allowing for country effects, they classify people as poor, average,
and rich through self-categorization. In the survey questionnaire, people classify
themselves as poor, average, or rich by their own standard. From this study, Easterly and
Fischer found evidence supporting the view that inflation is regarded as more of a
problem by the poor than it is by the non-poor and that inflation reduces the relative
income of the poor.
The concern over the role of inflation on poverty has, since then, been increasing.
A study by the Institute for Fiscal Studies (2011) finds that inflation hits poorer families
much harder than the rich in the UK. Similar concerns have been expressed for
developing countries as well. As the effect of inflation on poverty draws popular interest,
the issue has drawn the interest of researchers as well. Considering the seriousness of
the issue, a number of studies have been carried out at different times, in different parts
of the world, to understand the impact of inflation on poverty.
Cardoso (1992), Powers (1995), Ravallion (1998), and Braumann (2004) find that
there is a positive correlation between inflation and poverty. Chaudhry and Chaudhry
(2008) found that food price inflation increases poverty in Pakistan. While there are
studies supporting the view that there is a positive relationship between inflation and
poverty, there are studies which negate this view as well. For instance, Blank and Blinder
(1985) found no such relationship between inflation and poverty. Cutler and Katz (1991),
in contrast, found that an increase in inflation reduces the poverty rate in the United
States. Romer and Romer (1998) found that, under certain conditions, there is a negative
relationship between inflation and poverty. The findings, so far, are fairly discrete in terms
of their spatial coverage, and mixed in terms of conclusion.
2
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Given the scale of poverty in the world (according to the World Bank's 2008
estimates, nearly 22.5% of the world's population still lives below the international poverty
line of $1.25 per day, see figure 1) and the importance of poverty alleviation, I believe the
impact of inflation on poverty and development deserves more attention. I, therefore,
intend to study the relationship between poverty and inflation on a global scale.
Specifically, the research questions I aim to address are:
3
Texas Tech University, Shahidur Rashid Talukdar, August 2012
CHAPTER 2
LITERATURE REVIEW
I have divided the literature review into a number of subsections. First, I discuss
the literature relevant to the concept and definition of poverty. After introducing the
concept of poverty, I discuss the debate about the status of poverty in the world among
various agencies. Then, I discuss the literature on the effect of inflation on poverty,
followed by a brief description of inflation inequality and how it affects our understanding
of the overall impact of inflation on poverty. I sum up the literature review with a summary
of the available literature on the present topic.
There has often been a debate regarding what constitutes poverty and, as a
result, there are many definitions of poverty. One of the most widely accepted definitions
of poverty comes from the United Nations (UN). According to the UN (1998),
As most of the definitions of poverty are quite broad, there are many approaches
to measure poverty. For the ease of measurement and comparison, I shall consider the
income poverty and adopt the absolute approach of measuring poverty. The absolute
approach sets a threshold level of income or consumption expenditure (Citro and
4
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Michael, 1995). This threshold is widely known as the poverty line in development
literature. An individual or a family living with an income or consumption expenditure
below the set poverty line is designated as poor with reference to the poverty line
(Ravallion, 1992).
Different countries have different national poverty lines (Sachs, 2005). These
differences in the national poverty lines make it difficult to compare and analyze the
poverty statistics across countries. Hence, in order to be able to objectively calculate and
compare the poverty rates across countries, the World Bank has introduced the
international poverty line. In 2008, the World Bank came out with a revised figure of $1.25
per day adjusted for 2005 purchasing-power-parity (PPP) as the international poverty
line, which was previously set at $1.00 per day (Ravallion, Chen and Sangruala, 2009).
This means a person with a daily income or consumption less than $1.25 lives below the
international poverty line. The percentage of population living below the poverty line is
referred to as the poverty head count ratio or simply the head count ratio (UNDP, 2003).
This approach is very popular because of its objectivity and simplicity.
As it is difficult to find poverty data for a large panel from census dataset, I
use the World Bank estimates for poverty data. The poverty data is estimated by
POVCAL interactive software.1 The POVCAL is computer executable for calculating
poverty measures from grouped data. It is designed to be used as a reliable tool for
routine poverty assessment work. It uses sound and accurate methods for calculating
poverty and inequality measures with only a basic computer and any of the various types
of grouped distributional data typically available, often in published form. For further
details about the POVCAL software, see Ravallion and Huppy (1991), Ravaliion (1992),
Datt (1998), and Ravallion and Chen (2001). The World Bank, with the help of POVCAL
software, estimated that in 2008, there were almost 1.29 billion people living below the
international poverty line.
1
PovcalNet: An online poverty analysis tool, developed by the World Bank, updated version
released in February, 2012.
Link: http://iresearch.worldbank.org/PovcalNet/index.htm
5
Texas Tech University, Shahidur Rashid Talukdar, August 2012
60
50
40
30
20
10
0
1981 1984 1987 1990 1993 1996 1999 2002 2005 2008
Although, according to the figure above, the poverty has declined since 1981,
however, with 1.29 billion people or 22.5 % of the world's population living below poverty
the poverty line ($1.25 per day), as of 2008, the world poverty is a matter of grave
concern. Hence, the search for the determinants of poverty becomes important.
Although there are many other factors that may cause poverty, inflation is
regarded as an influential factor in determining the poverty. The problem of poverty
intensifies even more when the prices of commodities in general, and food in particular,
increase. Several arguments have been made in support of the view that inflation
increases poverty (see Cardoso, 1992; Powers, 1995; Ravallion, 1998; Braumann, 2004;
Chaudhry and Chaudhry, 2008). Inflation has, accordingly, often been labeled as the
cruelest tax on the poor. However, empirical findings on the effect of inflation on
poverty, after controlling for the rate of economic growth, are, in fact, mixed. Blank and
Blinder (1985) with data from 1959 1983 period, studied the relationship between
6
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Cardoso (1992) studied the effect of inflation on poverty between 1970 and 1990
in the Latin American countries. She argued that Inflation increases poverty in two ways.
First, the inflation tax can reduce disposable real income. Second, if nominal wages
increase less than the price of goods consumed by wage earners, workers' real income
will decline. She found evidence that in Latin America, inflation affected the poor through
inflation tax but the effect was very small. Higher rates of inflation had resulted in higher
inflation taxes but unless the inflation was extremely high (above 100%) this increase in
inflation tax was less than 1%. However, she showed that the main effect of inflation on
poverty was manifested through real wages. She found that accelerating inflation reduces
real wages and increases poverty. According to her results, real wages fall by 14 percent
when inflation doubles.
Romer and Romer (1998) studied the impact of the United States monetary
policy on unemployment, poverty and inequality. They found that regression of the
change in poverty on the unanticipated change in inflation produced a small and
insignificant coefficient. However, the relationship between the change in poverty and the
anticipated change in inflation was significant. The point estimate implies that an
anticipated increase in inflation of one percentage point is associated with a decline in
poverty of 0.2 percentage points. According to Romer, unanticipated inflation reduces the
real value of nominal assets and liabilities. It therefore causes real capital losses for
nominal creditors and real capital gains for nominal debtors. If the poor are net nominal
debtors, these effects benefit them. However, owing to the fact that the poor households
7
Texas Tech University, Shahidur Rashid Talukdar, August 2012
usually hold liabilities of small size, these benefits from unanticipated inflation are
insignificant.
Another finding of theirs, in this context, is very interesting. They find that an
expansionary monetary policy aimed at rapid economic growth is associated with
improved conditions for the poor in the short run, but prudent monetary policy aimed at
low inflation and steady output growth is associated with enhanced well-being of the poor
in the long run. Romer contends that inflation can harm the poor by reducing the real
value of wages and transfers. Another argument is that since inflation reduces the value
of the cash-holdings, it hurts the poor by making them poorer.
Erosa and Ventura (2002) observe that poor households hold more cash relative
to other nancial assets than rich households do. The poor then pays a disproportionate
share of the ination tax and are hurt more by ination.
According to Braumann (2004), the fall in real wages during inflation can also be
linked to increasing poverty in Latin America during the last two decades. From an
examination of the data for the period 1960 - 1997, he shows that poverty maxima
coincided with inflation maxima. The living standards of the poor were most hurt by
inflationary macro policies that intended to favor them. He, therefore, concludes that
fighting inflation might thus be an important step towards reducing poverty.
In the Indian context, Ravallion (1998) studied the impact of higher food prices
with survey data spanning 1959-1994. He found that there was strong positive correlation
between higher prices and poverty. Later, Datt and Ravallion (2002), using panel data on
poverty from Indian states, showed that inflation adversely affected Indias poor and
attributed this effect primarily to adverse shocks on the real wage of unskilled labor.
Although previous arguments and evidence tends to support the view that inflation
affects poverty positively, there are counter arguments to this. The UN Report on the
World Social Situation 2010, Rethinking Poverty, raises a number of interesting
questions: If inflation reduces real wages, then employment should rise, creating more
income-earning opportunities for workers. Therefore, the employment effect of inflation
(creating more jobs because of lower labor costs) can outweigh the real-wage effect
(lower income) on poverty. This is likely to be the case, as the inflation (real wage)
8
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Besides the direct effect of inflation on poverty, another issue that seems to have
a substantial impact on poverty is the inflation differentials across sectors. Inflation is not
the same in all the sectors and hence its impact on different section of population will also
be different. Given the heterogeneity of consumption pattern among various segment of
population, the effect of inflation will also be different in each section.
Son and Kakawani (2006), in the context of Brazil between 1999 and 2006, found
that if food prices increased by 1%, the poverty head count ratio would increase by
0.42%. If non-food prices increased by 1%, the head count ratio increased by 1.02%.
Thus, they found that if the overall prices increased by 1%, the poverty head count ratio
increased by 1.44%.
In the case of Indonesia, Sugema, et al. (2010) have found evidence that inflation
has a larger impact on poor households both in rural and urban area relative to non-poor
9
Texas Tech University, Shahidur Rashid Talukdar, August 2012
household. They also found that rural poor households are more vulnerable to economic
shocks, especially inflation. Furthermore, they find that price fluctuation on foods and its
products has a larger impact on poverty relative to non-food commodity.
A study by the Asian Development Bank (2011) noted that global food prices
increased by more than 30% in the first 2 months of 2011, and domestic food inflation
averaged about 10% for a number of regional economies. The study further found that a
10% rise in domestic food prices in developing Asia risks creating an additional 64.4
million poor people, or increasing the percentage of poor by 1.9 points. So a 30%
increase in global food prices would increase the percentage of poor by 5.7%.
10
Texas Tech University, Shahidur Rashid Talukdar, August 2012
CHAPTER 3
RESEARCH METHODOLOGY
In order to develop a general understanding of how inflation affects poverty, I form
a panel of 115 developing countries spanning the time period from 1981 through 2008.
However, within these developing countries, there are vast differences among them in
terms of their levels of income, inflation, consumption, and poverty (for a brief discussion
on the differences across countries, see Chapter 4 on Data). Given such differences
across countries, not much can be inferred from an attempt to study the effect of inflation
on poverty on a global scale. And studying just one country at a time contributes to the
country specific results but may not contribute much to the general understanding of the
effect. Keeping these points in mind, I first analyze the relationship between poverty and
inflation for all the countries and then group the countries into three separate groups: Low
Income Countries, Lower Middle Income Countries, and Upper Middle Income Countries.
For classifying the countries according to their income, I follow the World Banks
classification. In this method, economies are divided according to 2009 GNI per capita,
calculated using the World Bank Atlas method.1 The groups are: low income, $995 or
less; lower middle income, $996 - $3,945; upper middle income, $3,946 - $12,195;
and high income, $12,196 or more. A full list of all the countries according to their level of
income is included in the appendix.
11
Texas Tech University, Shahidur Rashid Talukdar, August 2012
The debtor country has to pay interest on the loan amount to the creditor country
or agency. The interest payment made to the creditor agency reduces the funds available
to the debtor country which the latter could otherwise utilize for its development. As a
result, the amount of external debt may affect the wellbeing of the citizens of the country.
However, the relationship between poverty and external debt may not be
straightforward (Shreideh, 2008). Maier (2005) found evidence supporting the view that
higher external debts are associated with negative effects on the levels of income of the
poorest 40% of the population. However, he cautioned that the empirical findings of the
impact of the external debt on poverty should be interpreted carefully as sensitivity
analysis showed inconsistent results. So, I include external debt as a control variable and
expect to find a statistically significant relationship between external debt and poverty
although at this point, I am unable to ascertain the directionality of the relationship.
Another factor that affects poverty is the state of governance. There are
arguments which support the view that democracy, as opposed to autocracy, helps in
reducing poverty and hence democracy should have a negative correlation with poverty.
Kamal (2000) argues that genuine democracy offers a unifying force, participatory
institutions, and a process that can bring different groups within a nation together to deal
with poverty. Thus, a negative relationship between democracy and poverty may be
expected.
However, there are counter arguments to this thesis, as well. For example,
Varshney (1999) shows that in some cases pure autocracies have succeeded in
12
Texas Tech University, Shahidur Rashid Talukdar, August 2012
eliminating poverty while democracies have not. So there may be some cases where
democracy can have positive correlation with poverty. I include the polity score
(difference between democracy score and autocracy score) as a proxy for the quality of
governance.
13
Texas Tech University, Shahidur Rashid Talukdar, August 2012
CHAPTER 4
DATA
Since poverty head count ratio is the dependent variable in this study, I have tried
my best to collect the absolute poverty data for the widest cross-section of countries
spanning the longest period of time. The World Banks POVCAL (February, 2012)
estimates poverty data for 139 countries for the period 1981 through 2008. POVCAL is
an online, interactive, computational poverty analysis tool. The estimation frequency is
three years. So the database has a total of 10 observations (including years 1981, 1984,
1987, 1990, 1993, 1996, 1999, 2002, 2005, and 2008) for most of the countries.
The size of the dataset is restricted by that of the poverty data. Ideally, therefore, I
should have 1390 (=139 x 10) observations. However, owing to the fact that some
countries have data for only one or two years, I exclude those countries from my study
and included only 115 countries in my panel. Thus, the actual number of observations
stands 1150 which is less than 1390 (See Table 1). Since some of the countries have
data for only a few of the variables, the panel remains unbalanced.
The World Bank (POVCAL) further classifies the countries into four income
groups: Low Income Countries, Lower Middle Income Countries, Upper Middle Income
Countries, and High Income Countries depending on their levels of income. The low
income group consists of 45 countries, the lower middle income group consists of 46
countries, and the upper middle income group has 24 countries in it. The high income
group contains just one country. As the high income group includes only one country
Slovenia, I exclude this lone-country group from my study.
Table 2 shows the number of countries I include in my study from each group
(income level) along with their mean level of income. In the Low Income group, there are
14
Texas Tech University, Shahidur Rashid Talukdar, August 2012
45 countries with an average per capita annual income of $1684. There are 46 countries
in the Lower Middle Income range. These countries have an average annual per capita
income of $4218. The 24 countries in the Upper Middle Income group have an average
annual income of $8425 per person. The average income for all the countries in the
dataset is about $4000 which is close to the average annual per capita income of the
Lower Middle Income group. (At the end, in the appendices, I enlist the entire dataset
with all the variables for all the countries and also the list of countries according to their
respective income categories).
I collect the poverty data from the World Bank's poverty estimates released in
February 2012. The poverty estimates have been made using POVCAL - an online
poverty analysis tool. POVCAL is an interactive computational tool that allows one to
replicate the estimates made by the World Bank's researchers in assessing the extent of
absolute as well as relative poverty in the world. For my study, I use the poverty head
count ratio (percentage of population living below the international poverty line) as the
measure of absolute poverty. In February 2012, the World Bank released a major update
of the developing world's poverty estimates for 1981-2008. For exact methodology of this
estimation see Ravallion and Chen (2002), Ravallion and Chen (2004), Ravallion and
Lokshin (2006), Ravallion and Chen (2008), Chen, Ravallion, and Sangruala (2009).
The new poverty estimates combine the purchasing power parity (PPP) adjusted
exchange rates for household consumption data from more than 850 household surveys
across the developing countries. Over 1.2 million randomly sampled households were
interviewed for the 2008 estimate, representing 90 percent of the population of the
15
Texas Tech University, Shahidur Rashid Talukdar, August 2012
developing world. Covering 139 developing countries and a span of 27 years (with 10
observations for each countries at 3 year intervals), this is one of the largest dataset
available for measuring the absolute poverty at present. From this pool, I form a panel
comprising of 115 developing countries from around the world and 10 periods.
By looking at the poverty data, we can see that for most of the countries, the
absolute poverty has fallen over the period between 1981 and 2008 (see Figure 1, which
shows the fall in aggregate poverty level). However, there is a high degree of
heterogeneity among countries with respect to their poverty status. Table 3 shows the
descriptive statistics of the dependent and independent variables.
%Change
POPBPL in CPI CPI GDPPC INTPMT SECSCHENR POLITY
The end of the year CPI data for all countries (for which I have poverty data from
the POVCAL dataset) was available from the International Monetary Funds (IMF) World
Economic Outlook (WEO) report dataset. I adjust the base year to 2005 wherever it is
required. As the value of CPI widely varies across countries, it will be interesting to see
the percentage change in CPI, i.e. the inflation figures for countries. For empirical
analysis, I use log of CPI and not CPI as such. In the data pool, Armenia has
experienced the highest percentage change in inflation in the year 1993 which is a
catastrophic annual CPI-inflation of 10896%. The lowest inflation figure is recorded for
Vietnam which is -34%.
I use the purchasing power parity (PPP) adjusted real GDP per capita from the
same World Economic Outlook (WEO) dataset. The countries under consideration vary a
great deal in terms of their income (GDP per capita figures). Among the countries
included in my study, the Gambia had the highest per capita income of $36,244 in 2008
while Mozambique with GDP per capita of $172.5 in 1984 had the lowest income.
Although every country has experienced an increase in the GDP per capita, this increase
is not uniform.
As a proxy for external debt, I use the data on interest payment made to foreign
countries as percentage of the national GDP. For most of the countries the data was
available from the World Banks Global Development Finance (GDF) dataset. From the
interest payments made, it is apparent that Guyana had the highest external debt relative
to its GDP as its interest payment was roughly 22% of its GDP.
17
Texas Tech University, Shahidur Rashid Talukdar, August 2012
percentage of the population in the official age group corresponding to this level of
education. It provides information on the share of population with a level of skills deemed
to be necessary for significant developmental progress.
I use the polity score as a proxy for democracy and political stability. It examines
concomitant qualities of democratic and autocratic authority in governing institutions,
rather than discreet and mutually exclusive forms of governance (Marshall, 2010). This
perspective envisions a spectrum of governing authority that spans from fully
institutionalized autocracies through mixed, or incoherent, authority regimes (termed
"anocracies") to fully institutionalized democracies. The polity score is defined as the
difference between the democracy score and the autocracy score of a country for a given
year. The polity score gives a measure of how democratic a country is against the other
extreme, that is, autocracy. The score is polity 2 series from the Polity IV project of the
Center for Systemic Peace, VA, USA. The polity score captures the regime authority
spectrum on a 21-point scale ranging from -10 (hereditary monarchy) to +10
(consolidated democracy) including zero (Marshall and Jaggars, 2010). The higher the
polity score, the more democratic a country is. If a country has a polity score of -10, it is
completely autocratic.
18
Texas Tech University, Shahidur Rashid Talukdar, August 2012
CHAPTER 5
EMPIRICAL MODELS
With the dependent and independent variables discussed in the previous section,
I construct and empirically estimate the coefficients of each of the independent variable
from the respective regressions.
POPBPLi,t = C + i,t LNCPI + i,t LNGDPPC + i,t INTPMT + i,t SECSCHENR + i,t
POLITY + i,t ...(1)
LNCPI = natural log of end of the year consumer price index (CPI),
of GDP,
of population,
19
Texas Tech University, Shahidur Rashid Talukdar, August 2012
POPBPLi,t = C + i,t LNCPI + i,t POPBPLi,t-1 + i,t LNGDPPC + i,t INTPMT + i,t
SECSCHENR + i,t POLITY + i,t (3)
where POPBPLi,t-1 denotes the one period lag of the dependent variable POPBPL.
It should be noted here that one period lag does not mean one year lag. Since the
frequency of the periods is 3 years, one period lag here means the value of the variable
three years ago. Once I set up the dynamic panel data model, I estimate the fixed effects
models for the dynamic case also.
20
Texas Tech University, Shahidur Rashid Talukdar, August 2012
CHAPTER 6
RESULTS
6.1. PANEL LEAST SQUARED MODEL
The static multivariate regression analysis results are tabulated below in tables 4,
5, 6, and 7. Table 4 enlists the regression results of the panel least squared model
described as Model 1 in section 4. Table 5 enlists the results from the country fixed
effects model, Table 6 records the results from the period fixed effects model, and Table
7 records the country and time fixed effects model. The Table 4 contains results from four
different regressions. The first regression (regression number 1, as indicated by the
column 1) in the respective tables analyzes data from all countries, the second
(regression number 2) analyzes data from low income countries, the third (regression
number 3) refers to lower middle income countries and the fourth regression (regression
number 4) draws data from the upper middle income countries. The first row gives the
column titles; the other columns in the table are self-explanatory.
The regression 1 (indicated in the first column of table 4) analyzes data from all of
the 115 countries together (as noted in the second column). It takes the poverty head
count ratio or the percentage of population living below the $1.25 per day poverty line
(POPBPL) as the dependent variable and natural logarithm of end of the year CPI
(LNCPI), natural logarithm of GDP per Capita (LNGDPPC), interest payment made on
external debt expressed as percentage of GDP (INTPMT), secondary school
enrollment ratio (SECSCHENR), and polity score (POLITY) in the fourth column as
regressors. The third column notes the model type; PLS indicates that the model is Panel
Least Squared model.
In the panel least squared regression (regression 1), the coefficient of LNCPI is
close to 0.67. This implies that, ceteris paribus, one percentage point increase in CPI is
associated with 0.67% increase in the POPBPL. i.e., an increase in the CPI by 1% can
result in an increase in the percentage of population living below the poverty line by
0.67%. The sign of this coefficient is as expected and the t-statistic is 4.65 (column G),
which is greater than t critical = 2.576 with 604 observations. The probability (prob = 0)
figure of zero means that there is 0% chance that this relationship has been recorded
21
Texas Tech University, Shahidur Rashid Talukdar, August 2012
purely by chance. The positive relationship between inflation and poverty is as expected.
This result is in agreement with the findings of Cardoso (1992), Powers (1995), Ravallion
(1998), Braumann (2004), and Chaudhary and Chaudhary (2008) that there is a positive
correlation between price level and poverty.
Researchers have given different explanation supporting the view that inflation
increases poverty. First, inflation hurts poor by making them poorer caused by erosion of
the value their cash-holdings. For instance, Erosa and Ventura (2002) find evidence that
poor households hold more cash relative to other financial assets than richer households.
Since inflation directly erodes the value of the cash holdings, the poor is hurt more than
non-poor. As a result, the poor becomes poorer with successive phases of inflation.
Another view is that inflation affects the poor directly through a decline in their real
wages owing to the short-run rigidity of nominal wages (Powers, 1995). Both these
explanation account for the sort run effects of inflation on poverty. Cardoso (1992) made
an even stronger assertion. She at first argued, and then provided evidence, that inflation
increases poverty by decreasing long term real wages.
Among the control variables, GDP per capita has the strongest correlation with
poverty. The coefficient on LNGDPPC = -11.60 indicates that one percentage point
increase in GDP per capita is associated with a decrease in the poverty head count ratio
by 11.6%. This is statistically significant and the magnitude of the t-statistic (= - 0.19771)
is again much larger than the critical value. Thus, I find evidence that economic growth
reduces poverty. This result is also according to my expectation. My finding conforms
with the findings of Barro and martin (1995), Todaro (1997), Norton (2002) that economic
growth reduces poverty.
22
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Maier (2005) found evidence supporting the view that higher external debt levels are
associated with negative effects on the level of the income of the poorest 40 percent.
However, he cautioned that empirical findings of the impact of the debt indicators on pro-
poor growth should be interpreted carefully due to inconsistent results of the sensitivity
analyses.
Although it is generally argued that higher external debt leads to higher levels of
poverty, but empirical findings suggest that the relationship is not straightforward. In fact,
there can be an argument in favor of the negative impact of external debt on poverty. For
instance, if the countries take loan from international agencies and other countries to fund
their poverty alleviation program and as a result if the poverty level goes down, then we
can find a negative relationship between external debt and poverty. But this is not
necessarily the only interpretation of this relationship. Although in the present case, I find
the evidence of a negative relationship between external debt and poverty but, later, in
the sensitivity analyses, the relationship becomes positive in some cases and becomes
statistically insignificant in some other cases. I shall discuss about the positive
relationship between external debt and poverty when such an occasion arises.
23
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Varshney (1999) shows that democracies have been slow and steady, but not
spectacular, in attacking poverty. In comparison, countries with long authoritarian rule
show no such pattern. Another study, Kamal (2000), puts forward the argument that
genuine democracy offers a unifying force, participatory institutions and a process that
can bring groups within nations (and also different nations) together to deal with poverty.
Thus, a small negative relationship between poverty and democracy may be expected. In
this case, I find a negative relationship between poverty and democracy which is
expected.
However, an antithesis of this observation is also not unheard of. Varshney (1999)
has shown that in some cases pure autocracies have succeeded in eliminating poverty
while democracies have not. So there may be some cases where a positive relationship
between democracy and poverty can exist. Later in subsequent analyses, I find the
evidence of positive relationship between democracy and poverty. I discuss the possible
reasons and findings of previous studies as the case arises.
The second regression on Table 4 details the results from analysis of the
relationship between the dependent variable and the independent variables in low
income countries. The results, in this case, are slightly different than those in the case of
all the countries combined. In the low income countries, the relationship between inflation
and poverty is positive. The coefficient on LNCPI is 0.17 which is small compared to the
coefficient in regression 1. The standard error is greater than the coefficient itself which
casts doubt on the relevance of the coefficient. The t-statistic is also small and the
probability that this relationship might have been observed by chance is quite high with a
P-value of 0.82. Although this coefficient is not statistically significant, the sign of the
coefficient is positive which is as expected. Even the R-squared (=0.44) is also low as
compared to the previous regression. This means that in low income countries, inflation
has a small positive relationship but the relationship is not significant. This apart, the
coefficients on LNGDPPC, INTPMT, and SECSCHENR have negative signs. That is, an
increase in GDP per capita reduces poverty and so does the interest payment and
secondary school enrollment. The sign of the coefficient on INTPMT is not as expected.
However, the coefficient on INTPMT is not statistically significant at 5% or even 10%
level.
24
Texas Tech University, Shahidur Rashid Talukdar, August 2012
25
Texas Tech University, Shahidur Rashid Talukdar, August 2012
The coefficient on POLITY in this case is positive. That means a countrys shift
from a relatively authoritarian regime to a relatively more democratic form of governance
increases poverty. This observation seems to contradict the previous observations.
However, the relationship between democracy and poverty may not always be
straightforward. For example, Bardhan (2001) notes that the empirical literature on
democracy and development is rather unhelpful and unpersuasive. He mentions in his
article Democracy and Development: A Complex Relationship that, the three surveys of
the empirical literature that he has seen come out with three different conclusions: One
by Sirowy and Inkeles (1991) is supportive of a negative relationship between democracy
and development; one by Campos (1994) is of a generally positive relationship; and the
one by Przeworski and Limongi (1993) is agnostic. I have already discussed the possible
explanations in favor of a negative relationship between poverty and democracy in the
previous section. In this article, Bardhan attributes the reason for a positive relationship
between democracy and poverty to the inefficiency of some of the democracies in the
developing countries.
Such a positive relationship between democracy and poverty may not be true for
all other countries but may be true for the low income countries. The reason is perhaps a
number of authoritarian regimes in some poor countries have been toppled and
democracies have been established or just a relative shift of regime characteristics has
taken place. However, this change has not resulted in reduced levels of poverty in those
countries. Hence, we see the positive relationship between polity score and the level of
poverty.
Overall, out of five independent variables, four have signs on the coefficients
similar to the previous regression. So the coefficients of the independent variables, in low
income countries, do not necessarily confirm or wholly contradict the findings in the case
of all the developing countries combined. Since the R-squared is low (= 0.44), the
independent variables can only explain 44% of the variation in the dependent variable.
From regression 3 we can see the relationship between the poverty head count
ratio and the independent variables in the lower middle income countries. Log of CPI has
a positive relationship with the poverty head count ratio. This implies that in lower middle
income countries also, an increase in the price level increases the percentage of
26
Texas Tech University, Shahidur Rashid Talukdar, August 2012
population below the poverty line. The coefficient on LNCPI is statistically significant at
1% level. The coefficients on other control variables, except INTPMT, have the expected
negative signs and are statistically significant as well. However, the estimated coefficient
on INTPMT is not statistically significant and the coefficient on POLITY is statistically
significant at 3% level. The low value of R-squared (= 0.38) shows that only about 38% of
the variations in the dependent variable can be explained by the independent variables.
Overall, the results are similar to those in the previous cases.
Regression 4 on Table 4 shows the relationship between the poverty head count
ratio and the independent variables for the upper middle income countries. The
coefficients, except those on INTPMT and POLITY, have expected signs. The
relationship between CPI and the poverty level is a strong, positive, and statistically
significant at 1% level. Although the coefficient on the polity score is positive, which is
unexpected, but it is not statistically significant at even 10% level. I have explained before
the possible reasons that may lead to negative relationship between external debt and
poverty, which we observe in this case. The low value of R-squared tells us that the
model explains only about 36% of the variation in the dependent variable.
To sum up the first set of results, I can say that while CPI shows a positive
relationship with poverty level in all cases, GDP per capita and educational attainment
show negative relationship with poverty. The other two control variables, INTPMT and
POLITY are not consistent in their signs or in the levels of statistical significance.
However, out of the four regressions, only in the first case, a high R-squared (= 0.6858 )
was observed. In all other cases, the regressions had low explaining power as indicated
by their respective R-squared values.
One of the reasons for low R-squared could be the inter-country heterogeneity, as
I have included countries from all over the world with different socioeconomic
characteristics in my study. The same can happen because of period specific factors, as
well. As a result, some information might have been left uncaptured by the model. So I
consider to run the regressions with Country Fixed Effects, Period Fixed Effects, and both
Country plus Period Fixed Effects. Before I go on to run regressions with country and
period fixed effects purely on the basis of intuition, I perform a formal F-ratio test to see
whether fixed effects model is a better fit than a pooled cross-section model. This tests
27
Texas Tech University, Shahidur Rashid Talukdar, August 2012
the null hypothesis that there are no differences across countries or periods against the
alternative hypothesis that there are differences across countries or periods or both.
Under the null hypothesis, all the constant terms are equal whereas the alternate
hypothesis is that the constants are different across countries or periods. Next, I
construct an F-ratio test for testing the significance of fixed effects.
For the F-ratio test, I use the first regression (regression #1) without fixed effects
from Table 4 for all countries and regression # 13 from Table 7 with both country and
period fixed effects. I use the F-ratio as defined in Econometric Analysis by Greene,
William H., 6th Edition, page 197.
(RFE2 RPooled2)/(n 1)
F(n 1, nT n K) =
(1 RFE2)/(nT n K)
the number of periods covered, RFE2 is the R-squared from the model with fixed effects
and RPooled2 is the R-squared from the model without fixed effects.
I run three different types of fixed effects models. First, country fixed effects
model, then period fixed effects model, and then country and period fixed effects model.
Table 5 records results from the regressions with country fixed effects, Table 6 details the
results from period fixed effects, and Table 7 records results from the country and period
fixed effects together. Each table has four regressions for countries with different income
levels as discussed in the case of Table 4. The structure of the Tables 5, 6, and 7 is the
same as that of Table 4. For the interpretation of each column, see the description of
Table 4.
28
Texas Tech University, Shahidur Rashid Talukdar, August 2012
6.3. PANEL LEAST SQUARED MODEL WITH COUNTRY FIXED EFFECTS (CFE)
In this section, I shall discuss the results from country fixed effects model. At first,
I shall talk about the case with all the countries, then I shall discuss the results in the
context of Low Income countries to be followed by the results from Lower Middle Income
countries. Finally, I shall detail the results for the Upper Middle Income countries.
The regression 5 on Table 5 records the estimation of the country fixed effects
model for all countries. The coefficient on LNCPI is strong, positive, and statistically
significant at 1% level. The coefficient is stronger than it was without the country fixed
effects. Apart from this, all other coefficients have negative signs. This is consistent with
the model with no fixed effects. However, except for the coefficient on LNGDPPC, which
is significant at 1% level, the other coefficients lose their levels of statistical significance.
While the coefficient on SECSCHENR is significant at 10% level, the coefficients on
INTPMT and POLITY are statistically insignificant. The increased value of R-squared
(=0.94) shows the models explaining power has increased with the introduction of
country fixed effects.
From regression 6 (Table 5), we observe that the coefficient on LNCPI (= 2.85) is
positive, much stronger than before, and most importantly, it is statistically significant.
This implies that with one percentage point increase in CPI, the percentage of population
living below the poverty line increases by 2.85%. This confirms that even in the case of
low income countries, an increase in the price level increases the poverty head count
ration. Among the control variables, the coefficient on LNGDPPC has the usual negative
sign and is statistically significant at 1% level. Although coefficients on POLITY and
SECSCHENR maintain the usual negative sign, they become statistically insignificant.
The coefficient on INTPMT changes its sign from negative in the previous models to
positive in this model, which is as per my expectation but it is statistically insignificant.
29
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Table 5. Results from the Panel Least Squared Model (PLS) with Country Fixed Effects (CFE)
Dependent Variable: Population Head Count Ratio (% of Population Living Below the Poverty Line)
Number
of
Regression Income Model Independent Coefficient Standard Probab R- Durbin- Observati
Number Level Type Variable s Error t-statistic ility Squared Watson ons
5 All Countries PLS (CFE)* LNCPI 1.162411 0.288422 4.030237 0.0001 0.9399 1.0067 604
LNGDPPC -12.66681 1.468067 -8.62823 0
INTPMT -0.063278 0.247568 -0.2556 0.7984
SECSCHENR -0.092148 0.051825 -1.77805 0.076
POLITY -0.0442 0.095894 -0.46092 0.6451
6 Low Income PLS (CFE)* LNCPI 2.846934 0.531367 5.357758 0 0.8507 1.03909 251
LNGDPPC -21.96211 1.976445 -11.1119 0
INTPMT 0.230296 0.399798 0.57603 0.5652
SECSCHENR -0.048732 0.134235 -0.36303 0.717
POLITY -0.093014 0.123823 -0.75119 0.4534
Lower Middle
7 Income PLS (CFE)* LNCPI 0.652668 0.177218 3.682858 0.0003 0.90424 1.12484 254
LNGDPPC -8.126109 1.928995 -4.21261 0
INTPMT 0.055451 0.333021 0.166508 0.8679
SECSCHENR -0.133459 0.064308 -2.07533 0.0393
POLITY -0.037485 0.190366 -0.19691 0.8441
Upper Middle
8 Income PLS (CFE)* LNCPI 0.628462 0.224339 2.8014 0.0064 0.90637 1.22119 102
LNGDPPC -3.937402 1.44215 -2.73023 0.0078
INTPMT 0.166656 0.241483 0.690135 0.4922
SECSCHENR -0.083783 0.061186 -1.3693 0.1748
POLITY 0.005319 0.075106 0.070817 0.9437
30
Texas Tech University, Shahidur Rashid Talukdar, August 2012
The reason for which I expect to see a positive relationship between external debt
and poverty is that external debt, although viewed as an instrument for financing
economic growth, does not contribute significantly to financing economic development in
developing countries (Ayadi and Ayadi, 2008). Rather than help improve the economic
performance of a country, in developing countries, external debt becomes a burden on
the economy. Nakatami and Herera (2007) argue that external debt becomes a self-
perpetuating mechanism of poverty aggravation, work over-exploitation and a constraint
on development in developing economies. However, as the coefficient on INTPMT is not
statistically significant, I cannot make any conclusion, at this point, about the effect of
external debt on poverty. Overall, the increased value of R-squared (= 0.85) gives more
credibility to the model.
From regression 7 we find that the coefficient on LNCPI is positive and
statistically significant. The coefficient on LNGDPPC has the expected negative sign
and is statistically significant. The coefficient on INTPMT has the expected positive sign
and the coefficient on SECSCHENR has the expected negative sign but both these are
statistically insignificant. The coefficient on the POLITY score has positive sign but
statistically insignificant. The model has a high value of R-squared = 0.90 which means
that in the country fixed effects model, the independent variables explain 90% of the
variation in the dependent variable. Similarly, regression 8 analyzes the case of upper
middle income countries. The results from upper middle income countries are not
significantly different from that of the lower middle income countries.
From the period fixed effects model (#9) in Table 6 for all countries, I find that
CPI has a statistically significant and positive relationship with poverty while GDP and
secondary school enrollment both show consistent and statistically significant negative
relationship with poverty. Although the coefficient on INTPMT is negative, it is
31
Texas Tech University, Shahidur Rashid Talukdar, August 2012
So to test the significance of period fixed effects model, I perform the F-ratio test
comparing the present model with the initial model without any fixed effects. F(134,
1210) = 0.1430 which is much lower than the critical F-ratio (=1) with 134 and 1210
degrees of freedom. Hence, it appears that period fixed effects model has no particular
significance, in this case.
From Table 7, we can see that inflation has a positive relationship with poverty. A
coefficient of 1.27 on LNCPI means that if inflation increases by 1%, an additional 1.27%
people fall below the poverty line. GDP per capita has a negative relationship with
poverty. The coefficient (= -9.67) on LNGDPPC shows that if the per capita annual
income increases by 1%, poverty declines by 9.67% i.e. almost 10% people rise above
the poverty line who were previously below the line.. All other coefficients are negative
just as in the model with period fixed effects. However, we cannot infer much from these
coefficients, since none of these coefficients are statistically significant. The model has R-
squared = 0.94. This means that the independent variables can explain 94% of the
variation in the dependent variable.
The dynamic models, which include one period lag of the dependent variable as an
independent variable, are expected to capture those effects which have not been
otherwise accounted for. This way the models will have the least autocorrelation, if any,
32
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Table 6. Results from the Panel Least Squared Model (PLS) with Period Fixed Effects (PFE)
Dependent Variable: Population Head Count Ratio (% Population Living Below the Poverty Line)
Regression Income Model Independent Standard R- Durbin- Number of
Number Level Type Variable Coefficients Error t-statistic Probability Squared Watson Observations
9 All Countries PLS (PFE)* LNCPI 0.493902 0.22813 2.164999 0.0308 0.6907 0.2036 604
LNGDPPC -11.83688 3.523055 -3.35983 0.0008
INTPMT -0.54959 0.67049 -0.81969 0.4127
SECSCHENR -0.446293 0.106834 -4.17744 0
POLITY -0.097184 0.224358 -0.43317 0.6651
10 Low Income PLS (PFE)* LNCPI 1.168366 1.466905 0.796483 0.4266 0.46639 0.31085 251
LNGDPPC -6.597678 4.345877 -1.51815 0.1303
INTPMT -1.282053 1.371238 -0.93496 0.3508
SECSCHENR -0.439316 0.09223 -4.76325 0
POLITY 0.40948 0.288208 1.420781 0.1567
Lower Middle
11 Income PLS (PFE)* LNCPI 0.465184 0.30581 1.521155 0.1295 0.39362 0.17114 254
LNGDPPC -11.25981 3.522163 -3.19685 0.0016
INTPMT 0.392832 0.961903 0.408391 0.6834
SECSCHENR -0.272239 0.103041 -2.64204 0.0088
POLITY -0.18635 0.368171 -0.50615 0.6132
Upper Middle
12 Income PLS (PFE)* LNCPI 0.972672 0.377486 2.576708 0.0117 0.4312 0.3564 102
LNGDPPC -6.127081 2.089044 -2.93296 0.0043
INTPMT -1.601784 0.571874 -2.80094 0.0063
SECSCHENR -0.248357 0.04427 -5.61002 0
POLITY 0.260861 0.166708 1.564779 0.1213
* PLS: Panel Least Squared Model, CFE: Country Fixed Effects, PFE: Period Fixed Effects
Note: Estimates are Consistent with White's Heterogeneity Test.
33
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Table 7. Results from the Panel Least Squared Model (PLS) with Country and Period Fixed Effects (CFE + PFE)
Dependent Variable: Population Head Count Ratio (% Population Living Below the Poverty Line)
Number
of
Regression Income Model Independent Standard R- Durbin- Observ
Number Level Type Variable Coefficients Error t-statistic Probability Squared Watson ations
PLS
13 All Countries (CFE+PFE)* LNCPI 1.265397 0.383906 3.296113 0.0011 0.94255 1.01906 604
LNGDPPC -9.668968 2.024449 -4.7761 0
INTPMT -0.166065 0.316638 -0.52446 0.6002
SECSCHENR -0.063656 0.045833 -1.38888 0.1655
POLITY -0.01902 0.078948 -0.24092 0.8097
PLS
14 Low Income (CFE+PFE)* LNCPI 3.381219 1.103599 3.063812 0.0025 0.86463 1.0875 251
LNGDPPC -14.57654 4.401135 -3.312 0.0011
INTPMT -0.385703 0.57731 -0.6681 0.5049
SECSCHENR 0.063149 0.119303 0.529315 0.5972
POLITY 0.072478 0.127564 0.568172 0.5706
Lower Middle PLS
15 Income (CFE+PFE)* LNCPI 0.53749 0.318155 1.689397 0.0928 0.905907 1.091 254
LNGDPPC -9.75545 4.542895 -2.14741 0.033
INTPMT 0.149295 0.335323 0.445227 0.6567
SECSCHENR -0.142756 0.069834 -2.04423 0.0423
POLITY -0.112918 0.204252 -0.55284 0.581
Upper Middle PLS
16 Income (CFE+PFE)* LNCPI 0.137847 0.325757 0.423158 0.6735 0.92556 1.2046 102
LNGDPPC -9.828058 3.468942 -2.83316 0.006
INTPMT -0.03683 0.245805 -0.14983 0.8813
SECSCHENR -0.11478 0.062203 -1.84525 0.0693
POLITY -0.053184 0.102946 -0.51662 0.6071
* PLS: Panel Least Squared Model, CFE: Country Fixed Effects, PFE: Period Fixed Effects
Note: Estimates are Consistent with White's Heterogeneity Test
34
Texas Tech University, Shahidur Rashid Talukdar, August 2012
among the residuals. Further, the inclusion of lagged dependent variable as a regressor
would help us understand whether poverty at a given period is mainly because of
poverty in the previous period. Tables 8 through 11 record the results from regression
analyses with dynamic panel models. First, the results from the pooled regression is
presented in Table 8, then results from the dynamic panel regression with country fixed
effects is tabulated in Table 9, the regression with period fixed effects in Table 10 and
finally, Table 11 includes the results from regression analysis with both country and
period fixed effects.
From regression 17, it is clear that after the inclusion of lagged dependent
variable (LAGPOPBPL) in the set of regressors, the effect of inflation on poverty is
small but still positive, as the coefficient on LNCPI is merely 0.04. However, the
coefficient is statistically insignificant. The one period lagged value of the poverty
determines 0.85% of the variation in the poverty level in the current year. One percent
increase in the level of poverty in the previous period can increase the poverty level by
0.85% in the current period. The coefficient is statistically significant at 1% level. The
coefficient on LNGDPPC has a negative sign which is consistent with the previous
observations. The coefficient is statistically almost significant at 5% level. Educational
attainment has a small but negative effect on poverty which is statistically significant.
Besides these, the coefficients on INTPMT and POLITY have expected signs but they
are not statistically significant. The value of R-squared (= 0.936) for this regression is
also high which means that about 94% of the variations in the dependent variable can
be explained by the independent variables.
From regression 18 on Table 8, we can see that in the case of low income
countries, an increase in CPI affects poverty negatively but as the coefficient is not
statistically significant, we cant rely much on the coefficient. Lag of the poverty variable
has a strong, positive, and statistically significant impact on poverty. The relationship of
all other independent variables with poverty is similar to the case with all countries.
Although the coefficients on the other control variables are similar to those observed
previously, the negative coefficient on the inflation seems a bit puzzling. A negative
coefficient on LNCPI means that an increase in inflation is associated with reduced
35
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Table 8. Results from the Dynamic Panel Least Squared Model (DPLS)
Dependent Variable: Population Head Count Ratio (% Population Living Below the Poverty Line)
Regres
sion Income Model Independent Standard Durbin- Number of
Number Level Type Variable Coefficients Error t-statistic Probability R-Squared Watson Observations
17 All Countries DPLS* LNCPI 0.039339 0.07656 0.513829 0.6076 0.9368 1.8744 549
LAGPOPBPL 0.858748 0.021438 40.05728 0
LNGDPPC -1.400495 0.719884 -1.94545 0.0522
INTPMT 0.122117 0.176251 0.692859 0.4887
SECSCHENR -0.051342 0.015254 -3.36592 0.0008
POLITY -0.023485 0.042039 -0.55864 0.5766
18 Low Income DPLS* LNCPI -0.88917 0.559781 -1.58842 0.1136 0.826289 1.600671 230
LAGPOPBPL 0.803011 0.043594 18.42 0
LNGDPPC -1.771309 0.681516 -2.59907 0.01
INTPMT 0.055453 0.417317 0.132881 0.8944
SECSCHENR -0.070115 0.028257 -2.4813 0.0138
POLITY 0.015917 0.087007 0.182941 0.855
Lower Middle
19 DPLS* LNCPI 0.107548 0.079134 1.359072 0.1755 0.91058 2.6028 228
Income
LAGPOPBPL 0.879317 0.018125 48.51356 0
LNGDPPC 0.335204 1.160099 0.288944 0.7729
INTPMT 0.239162 0.151052 1.583314 0.1148
SECSCHENR -0.050944 0.025399 -2.00577 0.0461
POLITY -0.03403 0.063074 -0.53953 0.5901
Upper Middle
20 DPLS* LNCPI 0.046129 0.128254 0.359667 0.72 0.896152 2.3385 93
Income
LAGPOPBPL 0.879621 0.057583 15.27583 0
LNGDPPC -2.295673 0.506938 -4.52851 0
INTPMT 0.002661 0.147135 0.018086 0.9856
SECSCHENR 0.006393 0.017494 0.365443 0.7157
POLITY -0.018737 0.066727 -0.2808 0.7795
* PLS: Panel Least Squared Model, DPLS: Dynamic Least Squared Model
Note: Estimates are Consistent with White's Heterogeneity Test.
36
Texas Tech University, Shahidur Rashid Talukdar, August 2012
levels of poverty. This may have many different explanations. First, since poverty has a
negative correlation with income, one can argue that the governments in the low
income countries might be following expansionary monetary policies to boost economic
growth (Sullivan and Sheffrin, 2003) and reduce poverty and hence there is a negative
relationship between poverty and inflation.
Another possible explanation is that in the low income countries, along with other
countries, poverty is declining (Figure 1). More people can afford goods and services that
were previously out of their reach. So demand for essential commodities is increasing at
a higher rate in low income countries. Thus, reduced levels of poverty are pushing
demands for consumer goods and services up. As a result, the prices increase and
hence, a negative relationship between poverty level and inflation maybe observed in the
case of low income countries.
Apart from the theoretical arguments, there are empirical evidences supporting
the view that inflation can have a negative correlation with poverty, at least in particular
situations, if not in general. For example, Romer (1998) found evidence that in the case
of the United States, inflation has a negative correlation with poverty in the short run.
So there is a possibility that under certain circumstances, an increase in inflation may
be associated with a decline in poverty.
The results from regression 19 show evidence that in the case of lower middle
income countries, the price level has a positive relationship with poverty. However, this
37
Texas Tech University, Shahidur Rashid Talukdar, August 2012
time again, the coefficient is not significant. The relationship of other control variables,
except GDP per capita, with the dependent variable is similar to what is observed in the
case of all countries. One interesting observation is that, in this case, LNGDPPC has a
positive coefficient. This does not make much sense, although there are direct
evidences that in some cases poverty increases despite increase in GDP per capita.
For example, in the case of Albania, the poverty head count ratio was 0.17% in 1981
which grew 0.85% in 2005. The poverty nearly increased by 5 times in that period in
Albania.
Another country in the lower middle income range, Armenia, recorded a very
low poverty head count ratio of 0.87 in 1981 which increased to a maximum of 24.29%
in 1993 and then declined to 4.74% in 2005. The same trend is recorded in the case of
Bolivia where the poverty head count ratio was 1.99% in 1981 which rose to 26.27 in
1999 and slowly declined to 15.61% in 2008. However, in the period under
consideration, the GDP per capita also increased over time in all these countries.
Cases like these may have led to a positive correlation between GDP per capita and
poverty level. The reason behind such a positive relationship between income and
poverty may be due to increased inequality. To see the average effect of an increase in
GDP per capita, we need to consider the country fixed effects model which, I believe,
will explain the relationships more accurately.
The regression results for upper middle income countries show that inflation has
a small and positive but statistically insignificant impact on poverty. The relationship
between previous periods poverty level and current period poverty level is positive and
significant. The other coefficients, except SECSCHENR, show expected signs but the
coefficients are really small and statistically insignificant. So the overall learning from
the dynamic panel models is that inflation may have a positive correlation with poverty
but as the previous periods poverty has a significant bearing on the current years
poverty level, the correlation of poverty with inflation becomes almost insignificant.
However, as there is a great deal of heterogeneity among countries in this income
group, a country fixed effects model may do a better job of explaining the relationship
between the independent variables and the dependent variable.
38
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Table 9 records the results from regressions 21 through 24 which account for
country fixed effects models for respective regions. The foremost observation from this
analysis is that LNCPI has positive coefficients in all cases, and with country fixed
effects, the coefficients are significant in three cases, although it is statistically
insignificant in the case of lower middle income countries. Lag of poverty has a
significant positive coefficient in all cases. The coefficients on LNGDPPC and POLITY
are negative as expected. The other two coefficients are either inconsistent with the
previous results or statistically insignificant. Summarily, one can say that the dynamic
country fixed effects model reaffirms the broad conclusions drawn from the previous
models that increase CPI and the level of previous periods poverty affect poverty
positively, whereas an increase GDP per capita, in general, leads to a decline in
poverty level.
From the period fixed effects models for different regions (regressions 25 - 28),
the general observation is that CPI increase has a positive effect on poverty, with the
exception of Low Income countries. In the case of Low Income countries, the negative
coefficient on LNCPI indicates that poverty decreases as inflation increases. However,
the coefficient is not statistically significant. The lag of the poverty variable has a
positive correlation with poverty in all the cases. This implies that poverty in the current
period is mostly a result of poverty in the previous period. However, in the case of other
control variables, the results are either not as expected or they are not statistically
significant. From my previous experience, I have learnt that with this dataset the period
fixed effects model is not a good fit. So I move on to the country plus period fixed
effects.
6.8.7. Dynamic Panel Data Analysis with Country and Period Fixed Effects
Table 11 shows that after accounting for the country and period fixed effects,
the relationship between poverty and inflation is significant and positive in the case of
all countries together but negative in the case of low income countries, in particular. In
other cases, the coefficients are positive but not statistically significant. That means,
overall, an increase in the price level increases the poverty head count ratio. However,
39
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Table 9: Results from the Dynamic Panel Least Squared Model with Country Fixed Effects
Dependent Variable: Population Head Count Ratio (% Population Living Below the Poverty Line)
Number
of
Regression Income Model Independent Standard R- Durbin- Obser
Number Level Type Variable Coefficients Error t-statistic Probability Squared Watson vations
All
21 DPLS (CFE)* LNCPI 0.486573 0.181706 2.677809 0.0077 0.9593 2.1047 549
Countries
LAGPOPBPL 0.534189 0.069415 7.695547 0
LNGDPPC -7.103239 1.217766 -5.83301 0
INTPMT -0.142733 0.180676 -0.789993 0.43
SECSCHENR -0.023481 0.051859 -0.452785 0.6509
POLITY -0.173688 0.084797 -2.048284 0.0411
Low
22 DPLS (CFE)* LNCPI 1.567205 0.529492 2.959828 0.0035 0.896957 1.92696 230
Income
LAGPOPBPL 0.477473 0.080852 5.905551 0
LNGDPPC -14.64708 2.037102 -7.190155 0
INTPMT -0.32488 0.403193 -0.805766 0.4214
SECSCHENR 0.052106 0.137984 0.377627 0.7061
POLITY -0.239091 0.107486 -2.224385 0.0273
Lower
23 Middle DPLS (CFE)* LNCPI 0.219379 0.171368 1.280161 0.2021 0.933523 2.5427 228
Income
LAGPOPBPL 0.563072 0.126671 4.445139 0
LNGDPPC -2.690344 2.367944 -1.136152 0.2574
INTPMT 0.06623 0.218132 0.303622 0.7618
SECSCHENR -0.08242 0.050414 -1.634872 0.1038
POLITY -0.138709 0.119393 -1.161785 0.2469
Upper
24 Middle DPLS (CFE)* LNCPI 0.482027 0.131795 3.657399 0.0005 0.9244 1.9839 93
Income
LAGPOPBPL 0.412007 0.091543 4.500681 0
LNGDPPC -3.616913 0.905887 -3.992673 0.0002
INTPMT 0.146233 0.181648 0.805033 0.4235
SECSCHENR -0.033277 0.048603 -0.684667 0.4958
POLITY -0.059069 0.074639 -0.791397 0.4313
* PLS: Panel Least Squared Model, DPLS: Dynamic Least Squared Model, CFE: Country Fixed Effects
Note: Estimates are Consistent with White's Heterogeneity Test.
40
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Table 10: Results from the Dynamic Panel Least Squared Model with Period Fixed Effects
Dependent Variable: Population Head Count Ratio (% Population Living Below the Poverty Line)
Number
Regres of
sion Income Model Independent Standard R- Durbin- Obser
Number Level Type Variable Coefficients Error t-statistic Probability Squared Watson vations
25 All Countries DPLS (PFE)* LNCPI 0.660704 0.249601 2.64704 0.0084 0.961159 2.1458 549
LAGPOPBPL 0.528258 0.06515 8.108341 0
LNGDPPC -2.979227 1.879349 -1.585245 0.1136
INTPMT -0.191592 0.231789 -0.826578 0.4089
SECSCHENR 0.004975 0.051039 0.097465 0.9224
POLITY -0.130868 0.071011 -1.842918 0.066
26 Low Income DPLS (PFE)* LNCPI -0.276808 0.730488 -0.378936 0.7051 0.833201 1.62755 230
LAGPOPBPL 0.798808 0.053762 14.85822 0
LNGDPPC -1.3568 1.583483 -0.856845 0.3925
INTPMT -0.315741 0.41467 -0.761428 0.4472
SECSCHENR -0.066213 0.054726 -1.209902 0.2276
POLITY 0.066166 0.121894 0.542813 0.5878
Lower Middle
27 DPLS (PFE)* LNCPI 0.11721 0.082886 1.414106 0.1588 0.915796 2.6306 228
Income
LAGPOPBPL 0.888865 0.019874 44.72513 0
LNGDPPC 0.649963 1.284164 0.506138 0.6133
INTPMT 0.231147 0.136372 1.694969 0.0915
SECSCHENR -0.044757 0.024112 -1.856224 0.0648
POLITY -0.023337 0.053632 -0.435135 0.6639
Upper Middle
28 DPLS (PFE)* LNCPI 0.025815 0.165053 0.156407 0.8761 0.916953 2.2966 93
Income
LAGPOPBPL 0.881657 0.063104 13.97145 0
LNGDPPC -1.450361 1.482947 -0.978026 0.3311
INTPMT -0.092258 0.168588 -0.54724 0.5858
SECSCHENR 0.001143 0.019452 0.058747 0.9533
POLITY 0.038217 0.079074 0.483304 0.6302
* PLS: Panel Least Squared Model, DPLS: Dynamic Least Squared Model, CFE: Country Fixed Effects, PFE: Period Fixed Effects
Note: Estimates are Consistent with White's Heterogeneity Test.
41
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Table 11: Results from the Dynamic Panel Least Squared Model with Country and Period Fixed Effects
Dependent Variable: Population Head Count Ratio (% Population Living Below the Poverty Line)
Regress Number of
Ion Income Model Independent Standard R- Durbin- Obser
Number Level Type Variable Coefficients Error t-statistic Probability Squared Watson vations
PLS (CFE
29 All Countries + PFE)* LNCPI 0.660704 0.249601 2.64704 0.0084 0.9611 2.145865 549
LAGPOPBPL 0.528258 0.06515 8.108341 0
LNGDPPC -2.979227 1.879349 -1.585245 0.1136
INTPMT -0.191592 0.231789 -0.826578 0.4089
SECSCHENR 0.004975 0.051039 0.097465 0.9224
POLITY -0.130868 0.071011 -1.842918 0.066
PLS (CFE
30 Low Income + PFE)* LNCPI 2.790081 0.724459 3.851258 0.0002 0.9068 2.0151 230
LAGPOPBPL 0.444839 0.080433 5.53057 0
LNGDPPC -6.38478 2.715784 -2.350989 0.0198
INTPMT -0.94673 0.475299 -1.991861 0.0479
SECSCHENR 0.139218 0.128554 1.082954 0.2803
POLITY -0.110539 0.087192 -1.267763 0.2066
Lower Middle PLS (CFE
31 Income + PFE)* LNCPI 0.306923 0.292128 1.050644 0.2949 0.93713 2.5957 228
LAGPOPBPL 0.592808 0.131511 4.507683 0
LNGDPPC -1.026925 5.222913 -0.196619 0.8444
INTPMT 0.120265 0.218647 0.550039 0.583
SECSCHENR -0.052103 0.060303 -0.86402 0.3888
POLITY -0.180315 0.105899 -1.702707 0.0904
Upper Middle PLS (CFE
32 Income + PFE)* LNCPI 0.17149 0.246897 0.69458 0.4899 0.938828 1.939347 93
LAGPOPBPL 0.424739 0.10228 4.152722 0.0001
LNGDPPC -6.296073 3.988461 -1.578572 0.1194
INTPMT 0.013987 0.254243 0.055013 0.9563
SECSCHENR -0.053694 0.050517 -1.062897 0.2919
POLITY -0.035597 0.103834 -0.34283 0.7329
* PLS: Panel Least Squared Model, DPLS: Dynamic Least Squared Model, CFE: Country Fixed Effects, PFE: Period Fixed Effects
Note: Estimates are Consistent with White's Heterogeneity Test.
42
Texas Tech University, Shahidur Rashid Talukdar, August 2012
percentage of population living below the poverty line decreases as inflation increases.
The level of poverty in the previous period is a significant determinant of poverty in the
current period. GDP per capita and polity score affect poverty negatively. The
relationship between external debt and poverty is not consistent. In some cases it is
positive while in other cases, it is negative. Educational attainment also fails the
robustness check.
6.9. SUMMARY OF THE RESULTS AND DISCUSSION
Table 12 summarizes the regression results from the entire study. The table at
first records the findings from analysis with all the countries and then from countries of
different income groups separately. The dependent variable is the percentage of
population living below the poverty line. In the first row, the first column in each case
enlists the independent variable, second (column) indicates that the coefficients in the
subsequent rows are from panel list squared models (PLS). The third column records
results from the country fixed effects model (PLS-CFE), the fourth column records results
from period fixed effects model (PLS-PFE), and the fifth column shows the results from
models with both country and period fixed effects (PLS-CFE+PFE). Similarly, the sixth,
seventh, eighth, and ninth columns respectively record the results from the dynamic
panel model all different specifications. Inside each little rectangle, the value on the top is
the estimated coefficient, the (*) on the superscript gives the level of statistical
significance, the value below the coefficient within the parenthesis ( ) is the standard
error, and X denotes the case where the coefficient is not available or applicable.
To sum up the results, in the case of all countries combined, inflation has a
positive relationship with poverty in all the cases. This result is similar to the findings of
Cardoso (1992), Powers (1995), Ravallion (1998), Braumann (2004), Chaudhary and
Chaudhary (2008), and Chani, et. al. (2011) that there is a positive correlation between
price level and poverty. The relationship between GDP per capita and poverty is negative
and is robust in all specifications. My findings fall in line with those of Barro and Martin
(1995), Todaro (1997), Norton (2002), to name a few. I find that quality of governance
has a negative relationship with poverty.This is also in agreement with previous works
such as Varshney (1999) and Kamal (2000) as discussed in the previous section.
External debt and educational attainment have negative relationship with poverty in
almost all cases. I have tried to explain the negative relationship between external debt
43
Texas Tech University, Shahidur Rashid Talukdar, August 2012
1
PLS: Panel Least Squared Model, CFE: Country Fixed Effects, PFE: Period Fixed Effects, and DPLS: Dynamic Panel
Least Squared Model
2
Numbers in the parentheses indicate the standard errors.
(***), (**), and (*) indicate that the coefficients are statistically significant at 1%, 5% and 10% levels respectively.
44
Texas Tech University, Shahidur Rashid Talukdar, August 2012
1
PLS: Panel Least Squared Model, CFE: Country Fixed Effects, PFE: Period Fixed Effects, and DPLS: Dynamic Panel
Least Squared Model
2
Numbers in the parentheses indicate the standard errors.
(***), (**), and (*) indicate that the coefficients are statistically significant at 1%, 5% and 10% levels respectively.
and poverty in the results section. Educational attainment shows positive relationship with
poverty level in a couple of cases. However, the coefficients are extremely small (of the
order of 0.004) and statistically insignificant. In the case of low income countries, inflation
shows positive relationship with poverty in 8 out of 10 specifications. The only two cases
in which inflation shows a negative relationship with poverty is the dynamic panel case
and the dynamic panel case with period fixed effects. However, with the dynamic panel
model, the coefficient on LNCPI is statistically significant only at 12% level of significance
and in the dynamic panel model with period fixed effect the coefficient is small (= 0.28)
as compared to other specifications and is statistically insignificant. Thus, the positive
relationship between inflation and poverty is not significant in either case. GDP per capita
shows a consistent negative relationship with poverty in all the specifications. The
coefficients on the other control variables are either inconsistent or statistically
insignificant or both. I have tried to provide some possible explanations in each case in
the result section.
In the case of lower and upper middle income countries, inflation shows a
consistent positive relationship in all the specifications. However, the GDP per capita
loses its consistency in sign in the case of lower middle income countries. Educational
attainment and polity show consistent negative relationship with poverty in the case of
lower middle income countries, however, external debt fails the robustness checks.
45
Texas Tech University, Shahidur Rashid Talukdar, August 2012
In the case of upper middle income countries, inflation and GDP per capita
consistently show positive and negative signs respectively. Other than these, the
coefficients on all other control variables are either inconsistent or statistically
insignificant or both. A striking observation comes from the dynamic models. In all the
cases where the lag of poverty (LAGPOPBPL) has been used as an independent
variable, it shows a strong, positive, and statistically significant coefficient. This means,
present poverty, inevitably, is a result of past poverty.
Most of the findings from this study are similar to what other researchers have
found in the past. So the question arises here is: What is new about this study? One of
the most salient features of my study is the dataset I have used. Poverty data is hard to
find for a large number of countries. So, in that sense, the first thing about my study is
that it makes use of a large dataset developed by the World Bank with the help of
POVCALNET - an interactive poverty analysis software. I cover 115 countries, for the
period 1981 - 2008. Since the most updated version of this dataset was released in
February, 2012, and I have used this in my study, it is one of the most up-to-date and
comprehensive studies, in terms of coverage, thus far, on poverty and inflation.
As the dataset has one of the widest cross-sections of countries, my result is also
highly generic. So, considering the case with all the developing countries, I find a positive
relationship between inflation and poverty. Thus, inflation, in general, keeping all else
equal, increases poverty. The second observation is that the effect of inflation on poverty
is not same in all the countries. For instance, although in lower and upper middle income
countries, inflation has a positive relationship with poverty, but in the case of low income
countries, under certain specifications (with dynamic panel model and dynamic panel
model with period fixed effect), inflation has a negative relationship with poverty.
There can be many explanations for the negative relationship between inflation
and poverty. I have discussed about the possible explanations in the results section. But,
to be certain about which explanation is correct, further study is required. However, as
the coefficient, in the latter case, is not statistically significant, I refrain from making any
conclusion. Other than this, inflation, evidently, affects poverty. The higher the level of
inflation, the higher the level of poverty.
The second important realization from my study is that GDP per capita is
negatively correlated with poverty. Higher levels of GDP per capita is associated, in most
46
Texas Tech University, Shahidur Rashid Talukdar, August 2012
cases, with lower levels of poverty, with the exception of lower middle income countries
under dynamic panel model and dynamic panel model with period fixed effects. However,
even in the cases where the relationship between income and poverty is positive, the
coefficients are not statistically significant. So growth, at large, is pro-poor. Another
important observation is that in some cases, external debt has a negative relationship
with poverty. This gives me an opportunity to think about the possible role of foreign debt
in fighting poverty. Some countries may be taking loans from international agencies and
other countries and utilizing this debt to fund their poverty reduction programs. This can
lead to a negative relationship between external debt and poverty. Thus, external debt, if
utilized properly, can help fight poverty.
Although I have carried out the analysis of the relationship between inflation and
poverty, along with control variables, I have had to come across many hurdles. One of
the most significant challenges of this study was to find a good dataset. Though some of
the variables were relatively easily available, finding poverty data for a large cross-
section of countries was hard, as many of the countries dont carry out survey or census
on an annual basis. As a result, I chose to take poverty data from the World Banks
(POVCAL) estimates. This data has a limitation: the data is not annual; instead, the data
series obtained from the POVCAL software is tri-annual. This is a very low frequency
data. However, since the data was available for a large number of countries, I chose this
dataset for the poverty data.
Another issue was finding the data for educational attainment. Ideally, I would
prefer to take percentage of high school graduates or college graduates as an indicator of
a countrys level of educational attainment. However, since such data was not available
for a large number of countries, I considered gross secondary school enrollment as a
proxy for educational attainment.
Despite these limitations, I was able to study the effect of inflation and other
control variables for a cross-section of 115 countries. From this study, I conclude that, in
general, inflation affects poverty positively. An increase in inflation pushes more people
below the poverty line. Further, this time I have studied the effect of inflation in countries
with different levels of income by grouping them separately into low income, lower middle
income, and upper middle income countries. By grouping the countries with respect to
their average income, I find that the effect of inflation in different income level is not
47
Texas Tech University, Shahidur Rashid Talukdar, August 2012
same. In the case of low income countries, the relationship between inflation and poverty
becomes negative when I account for the period fixed effects orin the dynamic panel
case. This opens up an opportunity to study the relationship in the case of low income
countries in further details. Especially, it would be interesting to study the causality of
such a relationship. Later on, I would also like to study, the effect of inflation in different
geographical regions.
Besides the unavailability of data, shortage of time is another factor. Since I had
to complete this study in a very short period of time, I could not study how or why inflation
affects poverty. From the existing literature, two factors seem to contribute to the
mechanism via which inflation affects poverty. One of the channels is erosion of the value
of the cash holdings and the other channel is the effect of inflation on the real wages.
Some of the studies find that short run rigidity of the nominal wages lead to a positive
relationship between inflation and poverty while others contend that inflation reduces real
wage in the long run and hence the effect of inflation is not a short run concern only. In
future, I would like to study the causality of the inflation and poverty relationship.
48
Texas Tech University, Shahidur Rashid Talukdar, August 2012
CHAPTER 7
CONCLUSION
To conclude, I can say that, in general, poverty is positively affected by inflation.
Keeping all else equal, an increase in inflation increases poverty. However, the effect of
inflation on poverty is not a very straightforward. Although in most cases, inflation has
shown positive relationship with poverty, but some of these results are not statistically
significant. In some cases, inflation has shown negative effect on poverty. In lower and
upper middle income countries, inflation has a positive relationship with poverty, but in
the case of low income countries, under certain specifications, inflation has a negative
relationship with poverty. However, the coefficient, in the latter case, is not statistically
significant. This apart, inflation, evidently, affects poverty. The higher the level of inflation,
the higher the level of poverty. Income, on the other hand, is generally negatively
correlated with poverty. An increase in the GDP per capita, leads to fall in the level of
poverty in absolute terms. However, the effect of income on poverty is not statistically
significant in all cases. But poverty in the previous period is a significant determinant of
poverty in the current period. This means that poverty itself is the cause of poverty in
most cases.
49
Texas Tech University, Shahidur Rashid Talukdar, August 2012
REFERENCES
Ayadi, F. S. and Ayadi, F. O. (2008). The Impact of External Debt on Economic
Growth: A comparative Study of Nigeria and South Africa. Journal of Sustainable
Development in Africa. Clarion, PA.
Barro, R., and Sala-I-Martin, X. (1995). Economic Growth. McGraw-Hill, New York.
Braumann, B. (2004). High Inflation and Real Wages. IMF Staff Papers, Vol. 51, No. 1,
International Monetary Fund. Washington, D.C.
Cardoso, E. (1992). Inflation and Poverty. NBER Working Paper. No. 4006. Cambridge,
MA.
Chani M. I., Pervaiz, Z., Jan, S.A., (2011). Poverty, Inflation and Economic Growth:
Empirical Evidence from Pakistan. World Applied Sciences Journal. Dubai, UAE. pp.
1058-1063.
Chauhary, T. T. and Chaudhary, A. A. (2008). The Effects of Rising Food and Fuel Costs
on Poverty in Pakistan. The Lahore Journal of Economics, Special Edition (September).
Lahore, Pakistan. pp. 117-138
Chen, S. and Ravallion M. (2008). China is Poorer than We Thought, but No Less
Successful in the Fight Against Poverty. World Bank Policy Working Paper. No. 4621.
Washington, D.C.
Chen, S. and Ravallion, M. (2008). A Dollar A Day Revisited. World Bank Policy Working
Paper. No. 4620. Washington, D.C.
Chen, S. and Ravallion, M. (2008). The Developing World is Poorer than We Thought,
but No Less Successful in the Fight Against Poverty. World Bank Policy Working
Paper. No. 4703.
50
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Datt, G. and Ravallion. M. (2002). Is India's Economic Growth Leaving the Poor Behind?
Journal of Economic Perspectives, Vol. 16, pp. 89 108.
Easterly, W. and Fischer, S. (2001). Inflation and the Poor. Journal of Money, Credit and
Banking, Vol. 33. No. 2.
Fengler, W. (2011). Taxing the Poor Through Inflation, Africa CanEnd Poverty,
Retrieved from http://blogs.worldbank.org/africacan/taxing-the-poor-through-inflation
Global Food Prices and Developing Asia. March 2011, Asian Development Bank,
Retrieved from http://www.adb.org/publications/global-food-price-inflation-and-
developing-asia
Gordon, D. (2005). Indicators of Poverty and Hunger. United Nations, New York.
Handbook on Poverty Statistics: Concepts, Methods and Policy Use, December, 2005.
United Nations, New York. Retrieved from
Http://Www.Adb.Org/Documents/Studies/Eco_Pov_Ban/Eco_Pov_Ban.Pdf
Howitt, P. (1990). Zero Inflation As A Long-Term Target For Monetary Policy. C. D. Howe
Institute, Toronto, Canada. Retrieved from
http://Www.Adb.Org/Documents/Studies/Eco_Pov_Ban/Eco_Pov_Ban.Pdf
Indicators for Monitoring the Millennium Development Goals (2003). United Nations,
New York.
Kamal, S. (2000). Democratization and Poverty Alleviation in South Asia. South Asian
Regional Workshop, Kathmandu, Nepal.
Ladd, Helen, F. (2011). Education and Poverty: Confronting the Evidence. Journal of
Policy Analysis and Management. Malden, MA.
Maier, R. (2005). External Debt and Pro-Poor Growth, Macroeconomics and Monetary
Economics. No. 0504031. Retrieved from
http://129.3.20.41/eps/mac/papers/0504/0504031.pdf
Marshall, M. G. (2010). Polity IV Project: Dataset Users Manual. The Center for
Systemic Peace, VA.
Marty, A. L., and Thornton, D. L. (1995). Is There A Case for Moderate Inflation?
Economic Review. pp. 27-37.
51
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Morley, S. and Alvarez, C. (1991). Recession and the Growth of Poverty in Argentina.
Inter-American Development Bank, Washington, D.C.
Nakatani, P. and Herrera, R. (2007). The South Has Already Repaid its External Debt to
the North, but the North Denies its Debt to the South. Monthly Review. Vol. 59. pp. 31 -
36.
O'grady , S. (2011). Soaring Inflation Hurts the Poor More. Institute for Fiscal Studies
(IFS) Report.
PovcalNet: an online poverty analysis tool, The World Bank, Retrieved from
http://iresearch.worldbank.org/PovcalNet/index.htm
Poverty and Inequality Analysis. The World Bank, 2010. Retrieved from
http://web.worldbank.org/wbsite/external/topics/extpoverty/0,,contentmdk:22569747~pag
epk:148956~pipk:216618~thesitepk:336992,00.html.
Ravallion, M. (1992). Poverty Freak: A Guide to Concepts and Methods. Living Standards
Measurement Papers, The World Bank, Washington D.C.
Ravallion, M. (1998). Reform, Food Prices and Poverty in India. Economic and Political
Weekly, Mumbai, India.
Ravallion, M. and Chen S. (2001). Measuring pro-poor growth. Policy Research Working
Paper Series 2666, The World Bank, Washington D.C.
Ravallion, M. and Lokshin, M. (2006). Testing Poverty Lines. Review of Income and
Wealth. Vol. 52, Issue 3, pp. 399-421.
Ravallion, M., Chen S., and Sangraula, P. (2009). Dollar a Day. The World Bank
Economic Review. pp. 163-184
Rethinking Poverty, Report on the World Social Situation (2010, United Nations, New
York.
52
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Romer, C. D. and Romer, D. H. (1999). Monetary policy and the well-being of the poor.
Economic Review, Federal Reserve Bank of Kansas City, issue Q I, pp. 21-49.
Shiller, R. J. (1996). Why Do People Dislike Inflation? Cowles Foundation for Research In
Economics, Yale University. Discussion Paper No. 1115.
Shraideh, N. (2008). Foreign Debt and Poverty. Social Forum, United Nations Office of
the Higher Commissioner for Human Rights, Geneva.
Son, H. H., Kakawani, N. (2006). Measuring The Impact Of Price Changes On Poverty.
International Poverty Centre, United Nations Development Program, Working Paper 33.
Sugema, I., Irawan, T., Adipurwanto, D., Holis, A., Bakhtiar, T. (2010). The Impact of
Inflation on Rural Poverty in Indonesiaan Econometrics Approach. International Research
Journal of Finance and Economics. Issue 58.
Sullivan, A., Sheffrin, S. Y. (2003). Economics: Principles in action. Upper Saddle River,
New Jersey. Pearson Prentice Hall. pp. 551
Varshney, A. (1998). Democracy and Poverty. The U.K. Department for International
Development, and the Institute of Development Studies, Sussex, U.K.
Wilson, L. (2011), Inflation, the Hidden Tax. The Center for Development. Retrieved from
http://drlwilson.com/articles/inflaton.htm
World Bank Sees Progress Against Extreme Poverty, But Flags Vulnerabilities (February
29, 2012), Press Release No:2012/297/DEC, Retrieved from www.worldbank.org on
2012-05-30.
World Hunger and Poverty Facts and Statistic, 2012. Retrieved from
http://www.worldhunger.org.
53
Texas Tech University, Shahidur Rashid Talukdar, August 2012
APPENDIX A
WORLD BANK INCOME BASED
CLASSIFICATION OF COUNTRIES
LOW INCOME COUNTRIES
54
Texas Tech University, Shahidur Rashid Talukdar, August 2012
Costa Rica
55
Texas Tech University, Shahidur Rashid Talukdar, August 2012
APPENDIX B
DATA SERIES*
LOW INCOME COUNTRIES
Country Year PopBPL GDPPC CPI CPI-Inflation SecSchEnr IntPmt Polity
Bangladesh 1981 44.18 331.261 34.028 19.481 15.35476 0.448016 -4
Bangladesh 1984 46.1 394.136 44.172 10.436 18.76351 0.598025 -7
Bangladesh 1987 47.85 440.744 60.711 10.163 17.97279 0.744642 -5
Bangladesh 1990 49.86 512.997 79.204 11.782 18.03683 0.637689 -5
Bangladesh 1993 50.68 594.82 86.413 5.444 0.472409 6
Bangladesh 1996 48.58 684.698 102.003 0.733 0.476198 6
Bangladesh 1999 54.49 787.812 125.52 3.073 46.79228 0.39889 6
Bangladesh 2002 52.85 920.411 135.637 3.833 50.57927 0.349482 6
Bangladesh 2005 50.47 1,134.02 163.168 7.064 45.71364 0.36831 6
Bangladesh 2008 46.62 1,413.98 204.904 6.031 45.41245 0.28722 -6
Benin 1981 53.47 613.143 32.901 0.795 15.94813 1.217771 -7
Benin 1984 63.03 639.484 35.461 10.265 20.67583 2.359411 -7
Benin 1987 64.5 661.725 35.535 -1.339 15.516 1.535147 -7
Benin 1990 65.98 736.688 37.093 1.112 0.923367 0
Benin 1993 65.8 829.446 41.264 2.888 0.65896 6
Benin 1996 61.75 914.079 70.067 6.914 0.821675 6
Benin 1999 67.29 1,015.97 73.363 -3.244 22.17601 1.032486 6
Benin 2002 47.33 1,135.33 83.476 1.222 27.08993 0.902254 6
Benin 2005 49.99 1,246.24 89.568 3.734 37.07148 0.37267 6
Benin 2008 44.79 1,420.00 103.908 9.897 0.622635 7
Bhutan 1981 47.35 539.701 109.717 9.717 0.001721 -10
Bhutan 1984 45.2 688.929 151.792 8.059 12.02269 0.024114 -10
Bhutan 1987 43.61 1,009.60 185.283 7.912 0.236106 -10
Bhutan 1990 51.01 1,316.27 241.132 9.418 0.763551 -10
Bhutan 1993 47.69 1,584.97 345 8.937 1.046692 -10
Bhutan 1996 47.69 2,048.84 447.264 8.316 0.826704 -10
Bhutan 1999 29.88 2,408.27 553.774 4.504 36.74025 0.562566 -10
Bhutan 2002 46.4 2,944.10 610.504 2.28 48.06704 0.384821 -10
Bhutan 2005 26.79 3,530.80 688.07 4.933 45.43338 0.274235 -6
Bhutan 2008 9.34 4,784.27 828.082 9.035 55.44877 3.087854 3
Burkina Faso 1981 73.02 375.83 37.571 6.543 2.56913 0.683448 -7
Burkina Faso 1984 77.26 406.215 47.067 7.493 3.10093 0.60292 -7
Burkina Faso 1987 72.49 519.512 47.223 2.981 4.97543 0.785626 -7
Burkina Faso 1990 61.92 569.489 48.356 -1.355 6.65229 0.501396 -7
Burkina Faso 1993 71.17 640.299 49.138 2.696 7.75343 0.775964 -5
Burkina Faso 1996 66.96 744.357 70.503 6.931 0.679439 -5
Burkina Faso 1999 70.03 868.089 74.646 0.642 9.12025 0.683527 -4
Burkina Faso 2002 56.54 962.592 80.222 3.935 9.98406 0.509061 0
Burkina Faso 2005 55.04 1,097.46 87.06 4.472 13.35987 0.317423 0
Burkina Faso 2008 45.06 1,278.95 100.864 11.582 17.74582 0.2233 0
Burundi 1981 85.24 242.072 13.433 10.986 2.83161 0.442893 -7
Burundi 1984 88.31 261.097 19.085 24.966 2.99239 0.971135 -7
Burundi 1987 87.03 315.92 19.766 4.985 3.83622 1.610714 -7
Burundi 1990 84.49 355.682 25.829 10.732 4.80301 1.259117 -7
Burundi 1993 84.24 365.316 33.467 15.532 6.56348 1.398228 0
Burundi 1996 85.91 299.52 62.001 37.384 1.123929 -5
Burundi 1999 86.43 314.828 93.823 20.71 0.99765 -1
Burundi 2002 85.9 317.899 115.177 3.529 1.1061 2
Burundi 2005 81.32 335.097 143.853 0.957 13.79438 1.576789 6
Burundi 2008 80.58 390.241 226.707 25.664 18.21889 0.701623 6
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
56
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
1 2
GDP. CAF: Central African Republic, CDR: Congo, Democratic Republic of.
57
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
58
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
59
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
60
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
61
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
62
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
63
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
64
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
65
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
66
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
67
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
68
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP PerCapita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
69
Texas Tech University, Shahidur Rashid Talukdar, August 2012
70
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP Per Capita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
71
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP Per Capita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
72
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP Per Capita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
73
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP Per Capita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
74
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP Per Capita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
75
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP Per Capita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
76
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP Per Capita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
77
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP Per Capita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
78
Texas Tech University, Shahidur Rashid Talukdar, August 2012
* PopBPL: Percentage of Population Living Below the Poverty Line, CPI: Consumer Price Index, GDPPC:
GDP Per Capita, SecSchEnr: Secondary School Enrollment Ratio, IntPmt: Interest Payment as fraction of
GDP.
79