determinants of basic needs fulfillment the case...
TRANSCRIPT
DETERMINANTS OF BASIC NEEDS FULFILLMENT
THE CASE OF PAKISTAN
Muhammad Azhar Khan
Reg. No. 100-SE/PhD/S05
Supervisor: Prof. Dr. Nasim S. Shirazi
Co-supervisor: Prof. Dr. Asad Zaman
School of Economics
International Institute of Islamic Economics
International Islamic University Islamabad
2012
DETERMINANTS OF BASIC NEEDS FULFILLMENT
THE CASE OF PAKISTAN
Muhammad Azhar Khan
Reg. No. 100-SE/PhD/S05
Supervisor: Prof. Dr. Nasim S. Shirazi
Co-supervisor: Prof. Dr. Asad Zaman
A Thesis Submitted to International Institute of Islamic
Economics(IIIE), International Islamic University (IIU) Islamabad,
Pakistan in partial fulfillment of the requirements for the award of the
Degree of Doctor of Philosophy in Economics
2012
DECLARATION
I, Muhammad Azhar Khan, Registration No. 100-SE/PhD/S05, a student of PhD in
Economics at International Institute of Islamic Economics (IIIE), International Islamic
University Islamabad (IIUI) Pakistan hereby declare that the work embodied in this
thesis entitled Determinants of Basic Needs Fulfillment The Case of Pakistan is the
result of original research and has not been submitted for any degree in any other
University or Institution.
---------------------
Muhammad Azhar Khan
i
ABSTRACT
This study investigates the impact of different socio economic indicators on basic
needs fulfillment in Pakistan. Basic needs gap index (BNGI) is dependent variable and
is used as proxy of basic needs fulfillment. Ordinary least squares (OLS) and two
different versions of empirical Bayes techniques have been applied on the time series
data of eight different regions of Pakistan with rural and urban bifurcation for the period
1979 – 2008. Significant factors are figured out of ten explanatory variables: per capita
income, per capita savings, remittances (domestic and foreign), human capital index,
household size, ratio of income of top 20 percent to bottom 20 percent , share of income
held by bottom 20 percent , higher education , unemployment, and dependency ratio.
Our final model comprises of the following four explanatory variables, per capita
income, human capital index, share of income held by bottom 20 percent, and
unemployment. It is found that per capita income and income held by bottom 20 percent
are highly correlated with BNGI in all the regions of Pakistan. It is also observed that
share of income held by bottom 20% is also a significant variable that affect BNGI.
Human capital index and unemployment showed mixed and sometimes contrasting
results for rural and urban regions. Income distribution is more uneven in urban areas
as compared to the rural areas. In the case of human capital, there is a considerable
difference in rural and urban areas of Pakistan.
Growth for the sake of growth is meaningless unless it reduces the miseries of
the masses. To make every person part of development process, it needs to ensure
that no one is underprivileged and marginalized. This can only be done when all the
basic needs of the individuals are met. To improve the indicators of basic needs
fulfillment it is important to improve the income share held by the poorest 20% people,
which is in accordance to the MDGs. This requires strong political will at the part of the
policy makers, the government officials, and the political parties.
ii
ACKNOWLEDGEMENTS
All the praise and thanks to Allah Almighty Who enabled me to accomplish this
challenging task; and may peace and blessings be upon His prophet Muhammad
(PBUH), the role model for the humanity.
Following the tradition of the prophet Muhammad (PBUH) in thanking people who
do us a favor, I would like to thank many people who contributed in my research work.
I owe a special note of gratitude to my supervisors Dr. Asad Zaman and Dr
Nasim Shah Shirazi whose invaluable support and consistent encouragement was a
source of motivation for me through the twists and turns of my belated research work.
Their unflagging enthusiasm for my work has been inspirational.
I would also wholeheartedly like to express my thanks and gratitude to Dr Eatzaz
Ahmad, Dr. Hafiz Muhammad Yasin, Dr Atiq ur Rehman, and Muhammad Siddique for
their support in my research work. I am also indebted to all my teachers, especially Dr.
Asad Zaman and Dr. Shaukat Niazi (Late). Thanks are also due to Muhammad Khan
(Deputy Secretary-KPK) for his invaluable support and prayers for my success.
My colleagues and friends, Muhammad Zahid, Mehtab Ahmad Abbasi, Mudassar
Nazir, Liaqat Ali, Shahid Razzaque, Qammar Abbas, Tahir Masood Bhatti, Khalid
Mahmood, and Malik Naseer Hussain deserve my heartfelt thanks. I am also gratified to
university staff especially Niaz Ali Shah, Tauqir Ahmad and Zaheer Ahmad. I am
thankful to HEC (Pakistan), and Higher Education Department (KPK) for extending the
financial support to pursue my doctoral studies.
I wish to express my appreciation and gratitude to my noble parents, especially
to my dear departed father. I also extend my appreciation to my wife and children for
their continuous support, help and prayers during the completion of this work.
iii
This Thesis is Dedicated to
My Family,
My Source of Inspiration
for Higher Studies.
iv
TABLE OF CONTENTS
Abstract i
Acknowledgements ii
Dedication iii
Table of contents iv-vi
List of Tables vii-ix
List of Figures x
List of Abbreviations xi
CHAPTER 1: INTRODUCTION 1-13
1.1 Background of the Study 5
1.2 Statement of the Problem 8
1.3 Objectives of the Study 9
1.4 Motivation for and Significance of the Study 10
1.5 Methodology in Brief 12
1.6 Organization of the Study 12
CHAPTER 2: LITERATURE REVIEW 14-25
2.1 Growth and Inequality 14
2.2 Growth and Poverty 17
2.3 The BNF Approach to Poverty 20
2.4 Empirical Studies on the BNF Approach 23
CHAPTER 3: THEORETICAL BACKGROUND 26-52
3.1 Growth, Development and Income Distribution 26
3.2 Poverty and Income Inequality 29
3.3 Different Approaches to Poverty 34
3.4 Measures of Poverty 44
3.5 Concluding Remarks 51
CHAPTER 4: DATA AND VARIABLES 53-114
4.1 Limitations of the Grouped Data 53
4.2 Data Sources 54
4.3 Variables Suggested for the Preliminary Model 56
v
4.4 Construction of Basic Needs Gap Index (BNGI) 87
CHAPTER 5: MODEL SPECIFICATION AND METHODOLOGY 115-146
5.1 Model Selection 116
5.2 Empirical Model 134
5.3 Estimation Approaches 139
CHAPTER 6: EMPIRICAL FINDINGS AND ANALYSIS 147-182
6.1 Rural Areas 149
6.2 Urban Areas 156
6.3 Overall Areas 162
6.4 Rural-Urban Analysis using Aggregate Prior 169
6.5 Sensitivity Analysis 178
CHAPTER7: FINDINGS AND CONCLUSION 183-193
7.1 Overview of the Study 183
7.2 Summery And Findings 185
7.3 Conclusions and Policy Recommendations 189
REFERENCES 194-209
APPENDICES 210-229
Appendix I-A 210
Appendix I-B 211
Appendix I-C 212
Appendix I-D 213
Appendix II-A 214
Appendix II-B 215
Appendix II-C 216
Appendix II-D 217
Appendix III-A 218
Appendix III-B 219
Appendix III-C 220
vi
Appendix III-D 221
Appendix IV-A 222
Appendix IV-B 223
Appendix IV-C 224
Appendix IV-D 225
Appendix IV-E 226
Appendix IV-F 227
Appendix IV-G 228
Appendix IV-H 229
vii
LIST OF TABLES
No. Table Description Page
4.1 Per Capita Income (Monthly) 62
4.2 Per Capita Savings (Monthly) 63
4.3(a) Percentage of Total Monthly Income by Foreign (F) Remittances
64
4.3(b) Percentage of Total Monthly Income by Foreign (D)
Remittances
65
4.3(c) Percentage of Total Monthly Income by Foreign (F+D)
Remittances
66
4.4 Household Size 69
4.5 Percentage Distribution of Earners (Both Sexes) by Degree
Level Edu:
70
4.6 Dependency Ratio 73
4.7 Labour Force Unemployment Rate (Un) 74
4.8 Share of Income held by bottom 20 Percent 78
4.9 Ratio of Income of top 20 % to bottom 20 % 79
4.10 Human Capital Index 84
4.11 Average Monthly Expenditure / Household (Rs) 92
1.12 Average Monthly Income / Household (Rs) 93
4.13 Percentage Distribution of Monthly Income Among
Households by Quintiles Pakistan
96
4.14 Percentage Distribution of Monthly Income Among
Households by Quintiles Punjab
97
4.15 Percentage Distribution of Monthly Income Among
Households by Quintiles Sindh
98
4.16 Percentage Distribution of Monthly Income Among
Households by Quintiles KPK
99
4.17 Percentage Distribution of Monthly Income Among
Households by Quintiles Balochistan
100
4.18 Poverty (Head Count Ratio)
104
viii
4.19 Construction of the BNGI 105
4.20 BNGI for all Regions of Pakistan 107
5.1 Correlations, Means and Standard Deviations Aggregate Rural and Urban Areas
118
5.2 Correlations, Means and Standard Deviations Overall Areas
119
5.3 Correlations, Means and Standard Deviations Rural Areas (four provinces)
121
5.4 Correlations, Means and Standard Deviations Urban Areas (four provinces)
122
5.5 Static Panel Model for Rural Areas 124
5.6 Static Panel Model for Urban Areas 125
5.7 Static Panel Model for Overall Areas 127
5.8 Static Panel Model for Aggregate Rural Urban Areas 128
5.9 Rural Areas (Auto Selected Model) 129
5.10 Urban Areas (Auto Selected Model) 131
5.11 Overall Areas (Auto Selected Model) 132
5.12 Aggregate Rural and Urban Areas (Auto Selected Model) 132
6.1a OLS and Empirical Bayes Estimates for Rural Punjab 151
6.1b OLS and Empirical Bayes Estimates for Rural Sindh 152
6.1c OLS and Empirical Bayes Estimates for Rural KPK 154
6.1d OLS and Empirical Bayes Estimates for Rural Balochistan 155
6.2a OLS and Empirical Bayes Estimates for Urban Punjab 157
6.2b OLS and Empirical Bayes Estimates for Urban Sindh 158
6.2c OLS and Empirical Bayes Estimates for Urban KPK 160
6.2d OLS and Empirical Bayes Estimates for Urban Balochistan 161
6.3a OLS and Empirical Bayes Estimates for Overall Punjab 163
6.3b OLS and Empirical Bayes Estimates for Overall Sindh 164
6.3c OLS and Empirical Bayes Estimates for Overall KPK 165
6.3d OLS and Empirical Bayes Estimates for Overall Balochistan 167
6.3e Summary of Results 168
6.4a OLS and Empirical Bayes Estimates for Rural Punjab (using
aggregate rural and urban prior)
170
ix
6.4b OLS and Empirical Bayes Estimates for Urban Punjab (using
aggregate rural and urban prior)
171
6.4c OLS and Empirical Bayes Estimates for Rural Sindh (using
aggregate rural and urban prior)
172
6.4d OLS and Empirical Bayes Estimates for Urban Sindh (using
aggregate rural and urban prior)
173
6.4e OLS and Empirical Bayes Estimates for Rural KPK (using
aggregate rural and urban prior)
174
6.4f OLS and Empirical Bayes Estimates for Urban KPK (using
aggregate rural and urban prior)
175
6.4g OLS and Empirical Bayes Estimates for Rural Balochistan
(using aggregate rural and urban prior)
176
6.4h OLS and Empirical Bayes Estimates for Urban Balochistan
(using aggregate rural and urban prior)
177
x
LIST OF FIGURES
No. Figure Description Page
4.1 Human Capital for Overall Regions 85
4.2 Human Capital for Rural Regions 86
4.3 Human Capital for Urban Regions 87
4.4 BNGI. Pakistan Rural and Urban 109
4.5 BNGI. Punjab Rural and Urban 110
4.6 BNGI. Sindh Rural and Urban 111
4.7 BNGI. KPK. Rural and Urban 112
4.8 BNGI. Balochistan Rural and Urban 113
xi
LIST OF ABBREVIATIONS
Abbreviation. Description
B20 Share of Income Held by Bottom 20 Percent
BNF Basic Needs Fulfillment
BNGI Basic Needs Gap Index
CZ William J. Carrington and Asad Zaman
D Domestic Remittances
EB Empirical Bayes
F Foreign Remittances
GDP Gross Domestic Product
HCI Human Capital Index
HCR Head Count Ratio
HE Higher Education
HP Cheng Hsiao and M. Hashem Pesaran
HS Household Size
IMF International Monetary Fund
KPK Khyber Pakhtunkhwa
MDGs Millennium Development Goals
OLS Ordinary Least Squares
PGI Poverty Gap Index
PSLM Pakistan Social and Living Standards Measurement Survey
Rem Remittances (Domestic and Foreign)
SPGI Squared Poverty Gap Index
Spc Per Capita Savings
T2B Ratio of Income Top 20% to Bottom 20 %
UNDP United Nations Development Programme
WTO World Trade Organization
Ypc Per Capita Income
1
CHAPTER 1
INTRODUCTION
In the second half of the 20th century massive economic growth and substantial
development took place in the world, but all the people did not equally benefit from this
development due to uneven distribution of wealth and income. Most of the developing
countries, particularly the poor ones, are facing various problems like macroeconomic
imbalance, poverty, extremely high dependence on agriculture, and uneven income
distribution among the classes. Macroeconomic imbalance includes high rate of
unemployment, inflation, negative balance of payments, exchange rate depreciation,
debt burden, low and inconsistent growth rates of gross domestic product, and low rates
of industrialization. The most important factors, relevant to this study, are the poverty
and unequal distribution of income which lead to socio-political and economic instability.
The rich and the affluent live a luxurious life whereas the poor are unable to meet their
basic needs. This inequality also leads to various socio-political problems.
According to the World Bank Report (2000), 2.72 billion people were living on
less than $2 a day in 1990. However, by 1998, the number of the poor people getting
less than $2 per day raised to 2.80 billion. This shows an increase of around 100 million
people living in adverse poverty within eight years despite an increase in income level
by an average of 2.5 percent per annum. According to a report of The United Nations
Development Programme (UNDP-2004), the high income group with 19 percent of the
world population received 84 percent of the global income in the year 2002. The report
indicates that the six richest countries in the world, with only 11 percent of the world
2
population, received two-third of the global income in the same year. It is also reported
that the total revenues of the world’s top 11 largest corporations during the year 2002
was about US$2 trillion, which is equivalent to twice the aggregated income of all the
low income countries and the ratio of income of the richest 20 countries to the poorest
20 countries is 40:1.
Milanovic and Yitzhaki (2002) observe that majority (97%) of the world poor
population is living in developing countries. The World Development Indicators
(WDI:2008) shows that in 2005, the highest quintile i.e. the richest twenty percent of the
world, accounted for 76.6% of total private consumption and the poorest quintile just
consumed 1.5%. Likewise, the poorest 10% of the world population accounted for just
0.5% of the total private consumption as against the wealthiest 10% people who could
consume around 59% of the aggregate. According to the World Bank Report (2008), the
fast emergent economies of Asia will be having more than 600 million rural population
living in extreme poverty conditions. Despite the fact that rural-urban migration is
prevailing, however, poverty seems to be the overriding phenomenon for quite a few
more decades. Moreover, the report says that agriculture will be the key factor for
poverty reduction in the 21st century, since about 75% of the poor people in the
developing countries are living in rural areas. About 2.1 billion of these people are living
on less than 2 Dollars a day and 880 million on less than 1 Dollar a day, all depending
on agriculture for their subsistence. According to world bank (2008), in 2004, 2.6 billion
people live on less than $2 per day, with three quarters of them in rural areas. In 2008,
1.29 billion people live on less than $1.25 per day which accounts 22.4 percent of total
population.
3
Given this situation as well as the rate of growth in population, the developing
countries ought to strive for high and sustainable rate of growth in aggregate output,
and to reduce macroeconomic imbalances and socio-economic inequalities, failing
which the problems of abject poverty and extreme inequalities will keep on worsening in
the developing countries leading to other socio-political problems including terrorism.
The much debated relationship between economic growth and inequality has
gained considerable space in development research. In his 1955 article, often referred
to as the inverted-U hypothesis, Simon Kuznets predicted that inequality is likely to fall
after having reached to some climax during the earlier period of development. This
rising trend in the inequality in the early periods followed by a downfall later on was
supposed to be associated with structural changes that took place because of changes
in technology and labor productivity as the laboring class will shift overtime from the less
productive traditional (agricultural) sector to more productive (industrial) sector. This
phenomenon of structural transformation and trickle-down effect has been highlighted
by Arther Lewis (1950’s). A fairly good number of studies supported these hypothesis
including Kravis (1960), Oshima (1962), Adelman and Moris (1971), Ahluwaila (1974)
etc. However, the inverted-U hypothesis tested for individual countries could not be
proved. Anand and Kanbur (1984), Field (1989), Deninger and Squire (1996) found no
evidence of such a curve to exist in individual countries.
Dagdeviren and Weeks (2001) stated that income redistribution is not a
necessary condition for poverty reduction. According to author, aggregate growth itself
is capable of reducing poverty, although redistribution can support achieve the target of
4
poverty reduction along with growth. The study suggested that growth combined with
redistribution would be more effective to reduce poverty. Deininger and Squire (1996)
presented the theme of income redistribution with a new data set.
Inequality in Pakistan has been a key issue in development strategies and social
reforms. However, the basic fallacy lies in measurement of inequality. In majority of
cases, the researchers lay stress on income or consumption inequalities in Pakistan
and a considerable literature on these lines prevails. Further, a number of different
measures for inequalities of income is available. Different researchers have therefore
employed different tools in their studies of income inequalities and poverty, keeping in
view the coherency of data sets available. The first attempt to measure income
inequalities in Pakistan is attributed to Haq (1964). The study analyzed income
inequalities within the highest income group, based on income tax data for 1948-49
through 1960-61. This was followed by a chain of studies, the worth mentioning, among
others, are Bergan (1967), Khandker (1973), Naseem (1973), Allaudin (1975) and Ayub
(1977). All these studies used the Gini coefficient as a measure of inequality.
An important problem associated with the estimation of income inequalities is the
selection of suitable units of measurement. Only few studies have preferred to use
‘household’ as a unit of measurement; while some others have used the ‘individual’ as a
unit. Another group of researchers have considered the ‘per adult equivalent
expenditure’ as the unit of measurement. These include Jehle (1990), Haq (1998),
Anwar (2003) and Jamal (2003) etc.
5
Keeping in view the importance of income distribution, this study aims at
evaluating the ‘basic needs gap’ for different regions of Pakistan, with rural-urban break
up. After the descriptive analysis, the study tries to estimate the possible determinants
of the basic needs gap index (proxy of basic needs fulfillment) in Pakistan. The earlier
studies available in this area of research discuss only the basic needs fulfillment in the
context of cross country analysis. However, there are some serious reservations about
determination and assessment of basic needs, since different countries having different
socio-political and cultural environments. Moreover, such empirical studies are very few
in number, obviously due to non-availability of uniform and consistent data sets for
developing countries. The present study takes the rural, urban and overall classification
of the four provinces of Pakistan based on the HIES (Household Integrated Economic
Survey) data from 1979 to 2008. Luckily, the data on the relevant variables are available
in Pakistan from HIES survey, and so we endeavor to conduct research in this rarely
explored area.
1.1 Background of the Study
The two concepts of economic growth and economic development are used as
synonyms for the general reader to perceive a steady increase in the availability of
goods and services measured by the GNP per capita. Although there is no hard and
fast line of demarcation to separate growth and development because boundaries of
both are overlapping, however both the areas differ in scope and coverage. Growth
theory focuses on the factors responsible for enhancement of national or per capita
income. The focal point of the theory of development, on the other hand, is the overall
6
socio-economic and institutional setup that advances overtime. Economic development
means improvement in education, health and other facets of human welfare. It also
encompasses the elements like life expectancy, infant survival rate, literacy rate, human
capital development and conservation of the environment etc.
The growth rate of income is vital to the course of economic development and it
plays the role of an engine. Following the momentum achieved through growth in
income, the development carriage moves forward. If the process of growth is persistent
and sustainable overtime, this leads to rebuilding and modernization of relevant
institutions in due course, improvement in democratic norms, equity in the distribution of
resources, decline in poverty and general progress in the standards of living. Topics like
income, savings, labor force, technological progress, capital formation and investment
etc. are discussed under the domain of growth. On the other hand, the impact of growth
on the whole socio-economic composition is discussed in the development literature. It
discusses the questions, for example, whether or not the GNP growth is equitably
passed on to different segments of the society; whether the standard of living gets
improved overtime or otherwise; what is happening to poverty and if there is any trade
off between growth and poverty; what are other factors of development beside growth
(social, political) etc. Thus economic development is much broader concept than
economic growth. Moreover, economic growth can be measured quantitatively whereas
only qualitative indices may be available to measure the pace of economic
development. Both the disciplines come closer to each other when distributional aspect
of income among different groups of society is discussed.
7
The basic needs fulfillment (BNF) approach to development is never opposed to
an expansion in GNP; rather its aim is to minimize poverty through the vehicle of
growth. However, certain doubts were raised and magnified in the literary circles as if
the BNF was possible at the cost of growth. However this line of thinking has been
historically and empirically proved wrong. The BNF approach may require a revision of
the development priorities, stringent redistributive measures, reallocation of resources
across production sectors, shifts in technological choices and readjustments in the
structure of the economy. However, an integration of all these components into a unified
and comprehensive development plan may be hazardous and challenging because their
net outcome may not always be welcomed by the powerful and political elites. This
argument explains the reluctance often shown by the policy makers in developing
countries while considering the BNF approach central to the development planning. The
much enchanted slogans and practices of globalization, privatization and liberalization
in transitional and other developing countries have failed in achieving the targets; rather
the economic conditions of the citizens of these countries have further worsened. The
BNF approach has been neglected and almost shown its exit from the literature.
According to Hasan, Z (1997), even the Islamic economists are confined to the juridical
aspects of BNF with little attention to its operational side. As this study is concerned
with the case of Pakistan, it covers not only descriptive analysis to appraise different
concepts of basic needs but also compares different regions of Pakistan in terms of
BNF indices and explores their linkage with other important variables.
8
1.2 Statement of the Problem
The idea that the basic needs of all human beings should be fulfilled before the
relatively less important wants of a few are met with, is the well established and widely
advocated principle of all major religions of the world. The BNF approach is concerned
with eradicating mass deprivation; an apprehension that has always been at the heart of
development programs.
In spite of the claims of curing poverty and inequalities, most of the developing
countries pursue the neoclassical agenda of free market (laissez-faire) that takes into
consideration efficiency, globalization, and privatization process blindly but ignores
equity in distribution. This strategy has promoted the dominance of poverty in most of
the developing economies. The problem can be tackled if the basic needs objective is
recovered and put at the centre of the development discourse. However to accomplish
this agenda, a strong will on the side of policy makers and political leaders is the
necessary condition needed.
This study aims at constructing and analyzing the Basic Needs Gap Index (proxy
of basic needs fulfillment) for different regions of Pakistan. The operational implications
of BNF and the possible and significant determinants of this target in Pakistan are also
explored. This is a need of the hour in the context of developing countries, including
Pakistan where the problem of poverty is severe. The empirical work done so far on the
issue and mechanism of BNF is very scarce. As discussed above, the main reason is
the non-availability of consistent data. To our good fortune, we have several editions of
9
the HIES at hand that contain comprehensive information about food, clothing, shelter,
health, education and many other economic and social variables.
The developed world is highlighting and struggling for the human rights; like the
freedom of expression, rights to have demonstrations and rights to detention etc;
whereas, developing and poor nations are facing acute deficiency of food, clothing and
shelter for masses, which should be at the top of development agenda in these
countries. The present study explores the basic needs fulfillment situation in different
parts of Pakistan and then steps forward to identify the factors responsible for fulfillment
of these needs, keeping in view the peculiar features of different regions. Finally the
study provides valuable information to the policymakers, think tanks and academicians,
public authorities and researchers in this area.
1.3 Objectives of the Study
Most of the prominent economists stressed on increase in GDP and it was
believed that the trickledown effect will take care of all problems, but some people like
Stiglitz (2002) are skeptical of this outcome. So there is a dire need to rethink alternative
solutions to address the problem of poverty. The basic needs approach to economic
development is one way of addressing the problem of poverty. The objective is to
enable people earn their livelihood with honor or obtain the basic necessities of life like
nutrition, housing, water and sanitation, followed by appropriate education and health
facilities so as to increase their productivity.
10
The study primarily aims to examine the factors responsible for poor performance
of BNF in different regions of Pakistan. The focus of the study is to investigate the
empirical relationship between basic needs gap index with per capita income, human
capital, status of employment and income distribution in Pakistan. The objectives of the
study are specifically explained as under:
To compute the Basic Needs Gap Index (BNGI) for Pakistan.
To explore and analyze the extent and severity of BNF situation in different
regions of Pakistan.
To investigate the empirical relationship between BNGI and per capita income,
unemployment, and the income share held by the bottom 20 percent of
population.
To compute and explore the impact of basic factors like human capital on
poverty, which is essential to get rid of the current mess that most of the
developing economies are facing.
1.4 Motivation for and Significance of the Study
When we see poverty amidst plenty, it is utmost necessary for a planner,
government in chair, political leader, and a researcher to think of seriously and to find
answers to this problem. For a student of Economics, this topic is much interesting to be
investigated. When we judge the situation with the usual yardstick of economic growth,
we see a spectacular, unprecedented, and appreciable success in GDP growth of many
countries after the World War II. However when we look at the world through the lens of
human characteristics, it is far less successful in elimination of poverty elimination and
11
reduction of income inequalities. The statistics reveal tremendous income disparities
across the nations and within different regions of nations. Different strategies to
accelerate development have been adopted overtime using the growth oriented
neoclassical approach, but is outcome is not encouraging. It is unfortunate that even
knowing the causes of failures, the governments and planners in the developing
countries follow the same policies time and again. Policies of the World Bank and IMF in
the form of Washington consensus and globalization are followed, which have failed in
most of the developing and transitional economies. So there is a dire need for the policy
makers in these countries to rethink and to focus on the reduction of poverty and
disparities. These burning issues inspired me to work on the topic of poverty and to
contribute something positive to the discipline.
On the one hand, there is unprecedented development in the field of science &
technology, trade & commerce, political liberty, human rights, longevity etc., yet
remarkable deprivation, destitution, and oppressions do exist around us. Given this
situation there is dire need to strive for high and consistent growth in aggregate output
on one hand and to reduce macroeconomic imbalances and socio economic
inequalities on the other, which are at the extremes in developing countries.
Hicks and Streeten (1979), while reviewing various social indicators, found that
the use of social and human indicators was capable to accelerate growth provided the
main targets include the fulfillment of basic needs. The present study focuses on this
possible approach to development, which seems to be the most important, live and
timely debate of the developing economies.
12
1.5 Methodology in Brief
The current study is directed to analyze empirically the issue of basic needs
fulfillment (BNF) using the data of different regions of Pakistan and the focus is to find
the factors responsible for this goal. Here the basic needs gap index is used as prime
variable and three different estimation techniques are used to find appropriate and
plausible parameters. These techniques include the ordinary least squares [OLS], the
empirical Bayes subdivided into Hsiao and Pesaran approved (2004) and Carrington
and Zaman (1994). The basic unit of analysis and comparison is the rural and urban
areas of the four provinces of Pakistan. Descriptive analysis of the regions is carried out
and we particularly focused on the spatial analysis to see the present and past state of
BNF and its linkage with the poverty.
The empirical Bayes technique is believed to give precise results particularly
when sample is small where the results of OLS analysis are imprecise. The present
study uses HIES data from 1979 to 2007-08.
1.6 Organization of the Study
The study is split into six chapters. The present introductory chapter is followed
by a review of the relevant literature that discuses the findings of different studies
focusing on the basic needs fulfillment approach. We present an overview the debate
on growth and development, which is followed by a brief account of studies on poverty
and income inequalities. The important empirical studies carried out on basic needs
fulfillment are also discussed. Chapter-3 provides the theoretical background to clarify
different approaches to handle the problem of abject poverty. Chapter-4 deals with data
13
and the description of variables and construction of different indices used in the
analysis. Chapter-5 deals with the empirical framework and methodology. It comprises
the discussion of the empirical model and different estimation approaches to the use of
empirical Bayesian technique. Chapter-6 is central and presents the empirical results
and analysis. These results are obtained by using three different techniques and the
findings are discussed group and region wise (rural, urban, overall, aggregate rural-
urban). The final chapter is reserved, as usual for conclusions and policy implications.
At the end, references are given.
14
CHAPTER 2
LITERATURE REVIEW
Equity in income distribution and fulfillment of basic needs have always enjoyed
high priority in major religions of the world. The prominent economists and social
philosophers have considered fulfillment of basic needs as an important factor of social
integration and a key to real economic development. Although, the issue of basic needs
fulfillment has lost its importance to some extent in the industrialized countries during
the 19th and 20th centuries of high growth and prosperity, however it remains a live topic
of discussion in the developing countries, which is emotionally argued at all forums and
investigated at academic sites. In this chapter, we briefly review the literature concerned
with different strategies of economic growth and development.
2.1 Growth and Inequality
Economic growth and development gained tremendous attention in 1950‟s. The
contributions of the Noble Laureate Arthur Lewis (1954) and others are worth
mentioning in this regard. The terms economic growth and development are used as
synonyms and perceived as continuous increase in the level of income along with an
uplift in the standard of living and development of the associated social, political and
economic institutions. Lewis (1954) contended that inequality was good for development
and economic growth, since saving is necessary for capital formulation, which in turn
leads to high employment, elimination of poverty and economic growth.
15
Another Nobel Laureate, Simon Kuznets (1955) argued from the historical facts
of European development that inequality worsened in the initial stage but later on
improved. The major deriving force behind this behaviour was presumably the structural
changes that occurred due migration of labour from poorer and less productive
traditional sectors to more productive industrial sectors. This line of thinking got
popularity under the title of inverted-U hypothesis, which was supported by a number of
studies conducted by prominent researchers like Harry Oshima (1962), Adelman and
Morris (1971), Felix Pankert (1973), Ahluwalia (1974), Robinson (1976) and Ram
(1988). However, the empirical evidence collected later on as well as the experience
gained during the later half of 20th century did not support the inverted-U hypothesis.
For instance, Ashwani Saith (1983), S. Anand and SMR Kanbur (1986), Gustav
Papanek (1987) argue against the existence of the hypothesis.
Subsequent research shows that there is no strong relationship between GNP
growth and the distribution of income. High growth rate does not necessarily worsen the
distribution of income as shown by Gustav Papanek and Oldrich Kyn (1986). However,
World Development Report 1991 provides evidence that higher growth is more often
associated with lower inequality. Greater equality leads to improved nutrition, more
employment and greater output growth, for instance the studies by Partha Dasgupta
and Debraj Ray (1987) and Roberto Parotti (1996) provide this argument. On the
contrary, Deininger and Squire (1996) attempted a comprehensive test and confirmed
that there was no systematic relationship between growth and inequality for individual
countries. On the hand, two classic technical articles by Abhijit V. Banerjee and Andrew
16
F. Newman (1993) and Oded Galz and Joseph Zaira (1993) address the mechanism by
which higher inequality may lead to lower growth or income.
Mahmood (1984) worked out income inequalities for urban and rural households
of Pakistan by using Gini coefficient, coefficient of variation, Theil entropy index and
Atkinsons indices. He used the time series data, ranging from 1963-64 to 1978-79, and
derived from household income and expenditure survey. The study revealed higher
intensity of income inequalities in urban area as compared to rural areas throughout the
study period. The study traced a declining trend in income inequalities up to the year
1968-69 in all regions. However rising trend persisted from 1970-71 onward in urban
Pakistan.
Kruijik (1986) estimated household income inequality in Pakistan, its all four
provinces and rural- urban break up of each province by employing Theil index. The
study used data for 1979. The coefficient of Theil index revealed the incidence of
income inequality was highest in KPK followed by Sindh, Punjab and Balochistan
respectively. According to the study urban income inequality was higher in urban areas
of all the provinces.
Ahmad et al, (1989) measured inequality in household income and expenditure
using Gini Coefficient, coefficient of variation, log-variance and Atkinson‟s indices and
different inequality aversion parameters for 1979 and 1984-85. According to the study,
inequality has increased from 1979-1984-85, but this increase was very trivial.
Jefri at (1995) analyzed inequality between urban and rural areas of Pakistan for
191 through 1979 by estimating Gini coefficient and income shares of the richest and
17
the poorest 20% households. The findings of study divulged that income inequality
improved slightly during 1979-88 but grew exorbitantly in 1990-91. According to study
inequalities in urban Pakistan have been persistent as compared to rural Pakistan.
Anwar (2003) analyzed inequalities for the period 1998-99 to 2001-02 by
incorporating household composition and using micro data of HIES. The study
measured inequality in per equivalent consumption expenditure for Pakistan as whole,
rural-urban cohorts of Pakistan as well as of provinces by estimating Gini coefficient.
The interesting feature of the study was adjusting data for household size and
composition by assigning weights to all members of household. The study observed
decrease in inequalities in three out of four provinces. According to study, Sindh and
overall Pakistan witnessed increase in inequality for the study period}.???
2.2 Growth and Poverty
The trickle-down effect in growth oriented approach didn‟t materialize because
the distribution side was altogether missing in the original policies. By now, there is
consensus among the researchers that mere emphasis on GDP growth is sufficient for
the process of some meaningful results. Instead, the reduction of inequalities, alleviation
of poverty and focus on basic needs fulfillment are the primary objectives of welfare and
development. The focus of research has shifted gradually during 1970‟s onwards while
measuring economic development. Anderson and White (2001) emphasize on the
pattern of growth and distribution. They derived data from 143 countries and found that
the growth effect dominates, but distribution also proved to be significant. In over a
18
quarter of cases, distribution played a stronger role than growth in increasing the
income for the poor.
Norman L. Hicks (1970) stressed on formulation of effective policies to promote
better distribution and reduce poverty since these policies are likely to stimulate growth
rather than retarding it. During 1980‟s and 1990‟s, the „Washington consensus‟
remained the advice in which fiscal strictness, privatization, and market liberalization
were the main ingredients. In 1980‟s, the economies of the Latin American countries
mostly suffered the grave problems of budget and BOP deficit, inefficient government
enterprises, protection policies, high inflation etc. However, the policies suggested by
the ‘Washington Consensus‟ were to be carried out in the developing countries where
the markets were imperfect and state intervention was enormous. As such the results
were not encouraging. Montek S. Ahluwalia, Nicholas G. carter and Hollis B. Chenery
(1979) also emphasize on the dimensions of poverty and inequality which formulating
policies for growth and development.
According to Afxentiou (1990), the unattended and uncontrolled poverty is a
threat to the social world fabric since this creates tensions among classes that may get
out of state control. The structure of asset ownership has led to more inequality of
income and wealth in many countries. It has prevented the poor to reap their shares
from growth benefits. The international organizations like ILO and the World Bank
advocate a re-orientation of development policies to deal with the problem of poverty
directly. Hence, emphasis on fulfillment of basic needs became a slogan in all
developing countries in late 1970‟s and early 1980‟s.
19
Joseph E. Stiglitz (2001) argued that the trickledown was a phenomenon that
never materialized in the democratic world. While quoting the example of USA, he noted
that although meaningful reductions in poverty cannot be attained without robust
economic growth, the converse is not true, i.e. it is not necessary that growth will
necessary benefit all. He showed that many third world and the transitional economies
were badly affected due to the haphazard process of privatization, liberalization and
stabilization initiated by the International Monetary Fund (IMF) and World Trade
Organization (WTO). He also advocated a reforms agenda for the betterment of life in
these countries where protests of trade unions and citizens from all walks of life were
going on against the strategy of liberalization.
Arvind (2006) notes that there is a wide realization among different segments of
the developed countries that India‟s economic growth has failed in reducing poverty. He
further states that poverty and inter-state disparities have increased over time. Even in
the presence of high growth rate, interregional disparities in income and poverty have
increased overtime.
The main focus of the capabilities approach as proposed by Amartya Sen (1995)
focuses on capabilities rather than consumption. He gave theoretical framework for the
provision of priorities in public policies. He suggests that social arrangement ought to be
evaluated keeping in view the level of freedom to enhance the valuable capabilities of
citizens. He refers to Alfred Marshal (1890) for the concept of human capital as the
personal wealth that includes all the energies, qualities and skills of the individual.
Likewise, Fisher (1906) had argued that labor participation in production was helpful for
20
enhancement of economic growth and poverty alleviation. He also draws upon Blaug
(1972) who stated that educated person was better in production process due to his
capabilities as compared to the uneducated person.
According to the UNDP report (2004), people are the real wealth of nations.
Indeed, the basic purpose of development is to enlarge human freedoms. The process
of development can expand human capabilities by expanding the choices that people
have to live full and creative lives. And people are both the beneficiaries of such
development and the agents of the progress and change that bring it about. This
process must benefit all individuals equitably and build on the participation of each of
them. Capability refers to a person‟s freedom to promote or achieve valuable
functioning.
2.3 The BNF Approach to Poverty Alleviation
According to Dudley Seers, (1969; 1972), due to the diminished faith in the
trickle-down effect, it was widely held that per capita GNP was not a proper yardstick to
measure development and an index of social welfare of the masses. As mentioned
above, the basic needs approach gained popularity and adopted as a strategy for
economic development during the late 1970‟s and early 1980‟s. The strategy focused
on essential needs necessary for long term physical well being, usually in terms of
consumption goods. In the words of Ghai (1977), efforts were made by the International
Labor Organization (ILO) at the „World Employment Conference 1976‟, which aimed to
redesign the global order through strategies that made the fulfillment of basic needs of
the poor the central focus for national and international efforts.
21
Streeten (1980) defends the Basic Needs Fulfillment approach by defining it as
the one that spells out (in considerable detail) human needs in terms of health, food,
education, water, shelter, transport and non-material needs like participation, culture,
identity and a sense of purpose in life and work, which interact with the material needs.
Further, Streeten says that “if we judge policies by reduction of suffering, the criterion of
basic needs fulfillment scores higher than that of reducing inequality, eliminating or
reducing unemployment or alleviating poverty.”
The basic needs approach is also used for the measurement of absolute poverty.
It attempts to define the absolute minimum resources necessary for long-term physical
well-being, usually in terms of consumption goods. consumption is considered a better
indicator than income as mentioned by A.V. Banerjee and E. Duflo (2006) , and Deaton
(2004). The poverty line is then defined as the amount of income required to satisfy
these needs .these personal needs include food, shelter, clothing, public service, health,
education, and safe drinking water etc. However, different agencies use different lists of
basic needs for the purpose of constructing development indices.
Hicks and Streeten (1979) are of the view that the use of social and human
indicators is most important supplement to GNP, especially if work on social indicators
is done in areas central to the basic needs approach, specially the human capital, which
could be instrumental in increasing productivity and growth in output. They are of the
view that GNP per head and related concepts are used to measure development
process in developing countries, whereas these are not sufficient indicators of
development and poverty elimination and fulfillment of basic human needs are goals
22
that should show up in a measure of development. Some early studies describing the
growth poverty nexus cast doubt on the capabilities of growth in reducing poverty,
unless expanding GNP is not complemented by social and human factors, it fails to
translate to the poor.
Some critiques also arise from various quarters of development economics. Main
attack on basic needs approach is that it is mighty difficult to identify a universal set of
basic needs which vary country to country and man to man. According to Weigel (1986),
a main question in the disagreement surrounding the basic needs approach to
economic development concerns the complicatedness of discovering a universal set of
basic needs which is capable of cross-cultural application. He further argues that the
apparent theoretical infeasibility of the basic needs approach stems from well-known
deficiencies in our current stock of economic, political and ethical paradigms,
particularly in the presence of rationality assumption.
Another critique states that by emphasizing activities that are essentially
consumption oriented, the basic-needs approach implies a reduction in the rate of
growth. However, Norman Hicks finds that the countries, which had performed well on
basic needs in 1960‟s had also shown growth rate above the average during the spell
1960-1973 and amelioration in basic needs during the same spell are also mutually
correlated with accelerated growth rates of GNP. Further, he ruled out the case that
higher growth in output was a cause of better basic needs conditions.
23
2.4 Empirical Studies on the BNF Approach
The existing literature in the area of basic needs fulfillment also includes a few
empirical studies, besides theoretical underpinning. These studies have been carried
out for different countries for different time spans and space and covering different
dimensions of basic needs. The wok of Stewart (1980) is worth mentioning where he
explored two observations that favored BNF. The first is the pattern of growth and the
resultant distribution of income that manifested an egalitarian labor deepening patterns
of growth, catering for the purchasing power of the poor to ensure sufficient nutrition
and others requirements for them. The second observation is on the nature of
government interventions that is necessitated in order to provide basic needs in the
form of both the levels of services and subsidies. The aim of such provision of services
and their distribution is to increase the income of the poor.
Hopkins and Hoeven (1983) investigated the role of economic growth and
income distribution in determining the basic need requirements. The study introduced
four basic determinates in this context. These included the level of per capita GDP,
number of years lapsed after independence, dependence on material exports and the
rate of economic growth and income distribution. According to the study, greater
number of years after getting independence could be associated with good
performance. On the other hand, dependence on mineral exports could be associated
with poor performance. The paper disclosed that both, rate of economic growth and
income distribution were insensitive to basic need performance. The study suggested
that policies promoting income distribution did not positively reduce a country‟s ability to
24
meet basic needs and that just acceleration of even appropriate economic growth would
not be dependable tool to solve the problem of poverty.
Hassan (1997) explored the correlation between basic needs and GNP per
capita for seven Muslim countries at three points in time (1987, 1990, and 1994). The
study used Basic Needs Gap Index (BNGI) as proxy for basic needs. This variable was
used as dependent variable. BNGI measures the extent by which the mean
consumption level of the poor, which equals their mean income, saving assumed to be
zero falls short of the mean national expenditure on basic needs. The study used
income, growth of income, net workers remittance and defense expenditures as
independent variables. Five basic needs food, clothing, shelter, Medicare and education
were included in the index. It measured the difference between the expenditure on
these basic needs by poor people and the mean national expenditure on them as a ratio
of the later on each year on a 0-1 scale. The study showed no correlation between the
BNGI and the GNP per capita or its rate of growth at any point in time. More over the
study ascertained that BNGI varied considerably over time and among countries. The
study found inverse relationship of BNGI (BNGI directly) with the pace of growth
between time splits 1987-90 and 1990-90. The study recommended multidimensional
sustained effort and political will and government intervention to carry into effect the
agenda of meeting basic needs.
Kipanga (2007) attempted to find the determinants of BNF based on data taken
from 27 developing countries from 1990 to 2002 in three groups as (1990-93, 1994-97,
1998-2002). The paper also used BNGI as the proxy of basic needs fulfillments. The
paper employed OLS pooled least squares techniques to find the determinations of
25
BNF. The study performed three types of analyses, firstly, the study made comparison
through the use of BNGI performance group averages, secondly cross country analysis
was done; and lastly the study employed multivariate regression test for variables.
Shirazi (1995), assessed the impact of various factors including Sadaqat on
poverty. Results conclude that the more Sadaqat to the poor, the less probability of a
household being poor. The Study also found that probability of a household being poor
was negatively related to the number of earners, educational level of the head and it
was positively related to the household size. Household in Punjab are relatively poor.
There are mixed feelings about the causes of poverty in developing countries.
Hassan (1997) and Navaratnam (2003) are of the view that affluent class is not ready to
sacrifice their growth prospects. They are least concerned with the living condition of
poor and lower middle income class. Most of the existing literature on the assessment
of basic needs fulfillment is concerned with reduction of poverty and inequality and very
few studies on this topic cover the basic needs approaches which provide basic tools to
analyze the relationship between BNGI and inequality.
In this research, the focus will be to analyze and compare the situation of basic
needs fulfillment in different parts of Pakistan. This study is very important because the
basic needs approach is envisaged as a dynamic instrument of development within the
framework of the current international economic order. This approach clearly tells
whether the benefits of development are passed on to the poor or otherwise. In
Millennium Development Goals (MDGs) redistribution of income along with growth is the
policy prescription for the developing countries.
26
CHAPTER 3
THEORETICAL BACKGROUND
This chapter is devoted to a brief review of different concepts of economic
development and their inter-relationships. The objective to provide a rationale for this
study while focusing on the Basic Needs Fulfillment approach (BNF).
3.1 Growth, Development and Income Distribution
Growth and development are the closely related terms that convey more or less
the same message to the general reader. However, where growth theory concentrates
on the factors responsible for uplifting the gross and per capita incomes, the theory of
development focuses on the overall socioeconomic structure and institutional set-up
that move ahead with the passage of time. The growth rate of income is central to the
process of economic development. The relationship between growth and development
resembles that of an engine and the carriage. Following the impetus of growth in
income/ output, the entire social and institutional structure of an economy begins to
improve in all directions. If the growth process sustains overtime, the social structure
moves gradually towards modernization, democratic attitudes, broadness in outlook
along with equity in distribution, reduction in poverty and general improvement in the
standard of living.
If we look at the very concepts of economic growth and development in
chronological order, it is revealed that they gained tremendous attention in 1950‟s after
the World War II. The cherished contributions of the Noble Laureate Arthur Lewis and
27
others are worth mentioning in this regard. Heavy emphasis was laid on economic
growth in the capacity of being a powerful tool for poverty eradication. The term
economic growth was perceived as continuous increase in the volume of goods and
services, measured by the GNP per capita. The underlying rationale was an
improvement in the quality of life through the trickle-down effect. Lewis (1954)
contended that inequality was good for development and economic growth, since only
rich could save and invest; and saving was necessary for capital formulation, which in
turn would lead to high employment, elimination of poverty and uplifting the standard of
living. The major driving force behind this phenomenon was presumably the structural
changes that occurred because of labor migration from poorer and less productive
traditional sectors to more productive industrial sectors.
Another Nobel Laureate, Simon Kuznets (1955) argued from the historical facts
of European development that with constant growth of per capita income overtime, the
inequality gets worsened in the initial stage but gets improved later on. This line of
thinking got popularity under the title of inverted-U hypothesis, which was supported by
a number of studies conducted by prominent researchers like Adelman and Morris
(1971), Ahluwalia (1974), Robinson (1976) and Ram (1988). However, the empirical
evidence collected later on as well as the experience during the later half of 20th century
did not support the inverted-U hypothesis. For instance, Deininger and Squire (1996)
attempted a comprehensive test and confirmed that there was no systematic
relationship between growth and inequality for individual countries.
28
The trickle-down effect in growth oriented approach didn‟t materialize because
the distribution side was altogether missing in that strategy. By now, there is sufficient
consensus among the development economists that the mere growth of income may be
necessary but never a sufficient condition and therefore a true representative of
development. Instead, alleviation of poverty, reduction of inequalities in income
distribution and the fulfillment of basic needs are the objectives that signify the extent of
development. The focus of research shifted gradually during 1970‟s onwards from
„GDP growth‟ to „growth with distribution‟ to „poverty reduction and basic needs
satisfaction‟, while suggesting strategies for economic development. Now the
distribution pattern and human capital receive more attention of researchers in the area.
Anderson and White (2001) express the pattern of growth and distribution in the
following words:
“The growth of income can be decomposed into a growth effect and a distribution effect.
Using data from 143 growth episodes, it is found that the growth effect dominates, but
distribution is important in a significant minority of cases. In over a quarter of cases,
distribution played a stronger role than growth in increasing the income for the poor.
Moreover, if there is no systematic relationship between growth and distribution, then it
is clearly better to have growth that is pro-poor rather than not in order to achieve
international poverty reduction targets.”
The purpose of sustainable economic development is to look for human
development and quality of environment. Sustainable development means use of
resources today by the people to satisfy their needs and improve their quality of life in
29
the present while safeguarding the ability of future generations to meet their own needs.
Current developments should not be on the cost of future generation. A better quality of
life means a higher standard of living which universally is measured in terms of income
level consumption expenditure and uses of available resources and technology. There
is a principle of equity Inherent in the concept of sustainable development. In order to
achieve economic and environmental goals, social goals – such as universal access to
education, safe drinking water, health care and economic opportunity – must also be
achieved, which increase quality of life and fruits should be distributed equally with the
future generations.
The senior vice president, chief economist of the World Bank and the winner of
the Nobel prize for economics 2001, Joseph E. Stiglitz in his book titled “Globalization
and its Discontents” –(2001) argues that, trickledown economics was never much more
than just a belief, an article of faith. He quotes the example of America:
“The economy grew in the 1980‟s and those at the bottom saw their real incomes
decline……..It is true that sustained reductions in poverty cannot be attained without
robust economic growth, the converse is not true: growth need not benefit all. It is not
true that „a rising tide lifts all boats‟ Sometimes a quickly rising tide especially when
accompanied by a storm that dashes weaker boats against the shore smashing them to
smithereens.”
3.2 Poverty and Income Inequality
Despite significant improvement and growth over the past half a century, extreme
poverty and gross inequalities remain widespread in the developing world. The fruits of
30
development have failed to reach the poor segments of the society, while the rich and
elite class is the prominent beneficiary. The reason is obvious. If the richer class owns
most of the productive assets (including agricultural lands), leading to higher income
and propensity to save and invest, they will naturally reap the benefits of growth through
the normal market mechanism based on towards functional distribution. The rents and
rentals of the physical assets as well as interest and profits of the financial capital and
business will accrue to this class that comprises the landlords, capitalists and business
class. The labouring class will reap the benefits of its time rented out in the form of
wages. The factor of production called „labour‟ is the weakest of all other factors since it
is human time that evaporates and cannot be stored for some useful employment.
Therefore, inequality is likely to increase with the process of growth and poverty has to
sustain overtime.
In fact for many countries, as noted by Todaro (2003), there is no particular
tendency for inequality to change much at all in the process of economic development.
Inequality seems to be a rather stable part of a country‟s socioeconomic make-up. This
inequality cannot be corrected peacefully and through a democratic process since it
involves a redistribution of productive assets. This can be possible only through forces
working outside the premises of market mechanism, through substantial upheavals and
social revolutions as in the case of former USSR, China, Japan, Tiawan and Korea.
It was asserted in the „Program of Action‟ at the Cairo International Conference
on Population and Development (1994) that “despite decades of development efforts,
both the gap between rich and poor nations and inequalities within nations have
31
widened over time;…….., and that the widespread poverty remains the main challenge
to development efforts”1.
Poverty is a multidimensional phenomenon and it requires multidimensional
policies and programs to eradicate it from this planet. Well-being of the individuals has
been conceptualized within multiple paradigms. It can be considered as command of a
person over resources and opportunities to earn. The people may be more
economically well off if they have more opportunities to manipulate and to enjoy the
commodities like food, clothing, and shelter, and other essentials of life. In this situation,
people will be less vulnerable to weather-shocks and income variations. Thus poverty
means either lack of command over commodities in general or a specific type of
consumption that deems essential for a reasonable standard of living in a society or lack
of an ability to function in a society.
Poverty is a social evil and it is generally understood as hunger, squalor, no
proper shelter, no access to education and health facilities, no security and access to
justice. Some time the lack of freedom of expression, the shortage of time and
restlessness is also considered as poverty. Wealth is useless if an individual cannot
purchase peace of mind and tiny pleasures of life. Defining poverty objectively is a futile
exercise because social researchers and reformers never agreed on a common line
which covers all the dimensions of poverty.
1 Referred to by Todaro & Smith: Economic Development – 8
th Edition (2003)
32
Poverty influences and is influenced by so many factors, in other words it is
caused by so many factors where as it also gives birth to many evils. Poverty prevails
unevenly among regions of developing world. It is observed that poverty is found more
in certain groups. Women and children are more victim of poverty. The incidence of
poverty is also observable among ethnic group and minorities. The term “Poverty„‟ in its
immediate sweep, is however, concerned with the poor and proletariat. Strictly
speaking, poverty may be defined as the proportion of people whose income falls below
a specific poverty line, generally known as head count ratio, the income gap. Poverty is
a deprivation of essential assets and opportunities to which every human being is
entitled. A deep postmortem of poverty reveals that it is not merely the command over
goods or services or calories intake; rather it is more complicated phenomenon. Thus
one can think of poverty from a non monetary2 perspective while measuring its different
components.
Another interesting characteristic of poverty is that sometime low incomes go side
by side with other forms of dispossession. For example in Mexico life expectancy for
poor 10 percent of population is 20 years less than for the richest 10 percent (Meir).
Poverty differs from inequality; poverty defines the absolute standard of living of a part
of society whereas inequality points to relative living standards across the whole
society. If a person has all income of nation, this is an example of maximum inequality.
2 Non Monetary approaches to measure poverty include Calories consumed per person per day, Food
consumption as a fraction of total Expenditure, Measures of outcomes rather than inputs and
Anthropological method.
33
In such case poverty would be high. Conversely, if all have same income there will be
no inequality, but poverty may be zero or maximum.
According to Afxentiou (1990), the unattended and uncontrolled poverty is a
threat to the social world fabric. Unattended poverty creates tensions among classes
and countries. The structure of asset ownership and its associated economic and
political power has led to more inequality of income and wealth. Supposedly, this
structure of asset ownership prevents the poor from benefiting from growth. The
objections against these economic disparities were raised by international organizations
like ILO and the World Bank. These advocates of re-orientation of economic
development stressed upon dealing with the poverty directly and gave this process the
name of basic needs. Hence, concentration on the basic needs became a slogan of
action, and this was referred to as a new theory of economic development by its
exponents while others termed it modestly as a new approach to development. In
subsequent period i.e. late 1980‟s, another approach Human Development also gained
popularity and given considerable weight.
3.2.1 Measuring Inequality
Inequality in income distribution can be measured in many ways. The „Personal
or size-distribution‟ looks at the gross income received by the individuals, irrespective of
whether it is earned or un-earned and the way it is received. A common method is to
divide the total population into successive quintiles or deciles according to ascending
income levels and then to estimate the proportion out of aggregate income each group
of people is deriving. The income shares of different groups are then mutually compared
34
to have an idea of inequalities in distribution. For instance, the income share of top 20%
of population (highest quintile) is compared with that of the bottom 20% (lowest quintile).
A more dis-aggregation, for instance deciles instead of quintiles, gives a clearer picture
of inequality in distribution.
A commonly used aggregate measure of inequality is the Gini Coefficient, which
is the ratio of the areas covered by the line of equality (hypothetical distribution) and the
Lorenz curve3 (actual distribution) to the total area beneath the line of equality, taken as
unity. The value of this coefficient lies between zero and unity. Thus a value nearer to
zero reflects more equity in distribution and a value nearer to unity shows the converse.
The coefficient of variations may also be used as a measure of inequality. Despite a few
short comings, the Gini coefficient is considered as the best indicator of macro-
inequality, particularly across countries and geographical regions.
3.3 Different Approaches to Poverty
A specific minimum level of income needed to satisfy the basic physical needs of
life (food, clothing and shelter etc) in order to ensure continued survival is defined as the
poverty line. Some problem however, arise when we recognize that the minimum
subsistence levels will vary from country to country and society to society, and even
individual to individual; reflecting different physiological, social and economic
requirements. One common methodology to avoid these problems is to define an
3 As referred to by Todaro (2003), Conard Lorenz, an American Statistician devised the convenient and
widely used diagram/curve in 1905 to show the relationship between population groups and their
respective income shares.
35
international poverty line and then attempt to estimate the purchasing power parity or
equivalent of that amount in terms of local currency. This amount was estimated by the
World Bank as US $ 370 per individual (1993 constant dollars) on annual basis, or 1.08
dollars per day. This became popular and commonly referred to as 1$ per day or
equivalent amount as a general yardstick.
This approach to poverty in terms of the basic needs fulfillment became popular
in the 1970‟s. The concern shifted from inequalities in distribution to the eradication of
absolute poverty, particularly by concentrating on basic human needs. It was generally
recognized that mass poverty can coexist with a high degree of equality and it is
possible to find ways and means to reduce mass poverty even if gross inequalities
prevail. Eradication of poverty through the provision of basic needs became the popular
political slogan4. Extensive Research started on this line of thinking with the pioneering
work of Norman Hicks and Paul Streeten (1979)5. In contrast to the common mass
perception of the basic needs as well as political slogan (food, clothing and shelter), the
economists have generally concentrated on identification of the measurable basic
needs to include food, education, health, clean drinking water and sanitation; on the
different indicators to be used for the assessment of basic needs and on the estimation
of cost involved for provision of these needs to the society at large.
4 Roti, Kapra, Makan was the slogan of PPP in 1970 elections of the „united‟ Pakistan.
5 “Indicators of Development: The Search for a Basic Needs Yardstick” – World Development Volume 7
(1979)
36
According to United Nations Development Program (2004), people are the real
wealth of nations. Indeed, the basic purpose of development is to enlarge human
freedoms. The process of development can expand human capabilities by expanding
the choices that people have to live full and creative lives. And people are both the
beneficiaries of such development and the agents of the progress and change that bring
it about. This process must benefit all individuals equitably and build on the participation
of each of them. Capability refers to a person‟s freedom to promote or achieve valuable
functioning.
The main focus of the capabilities approach expounded by Amartya Sen focuses
on capabilities rather than consumption. In his book “Inequality re-Examined” (1995), he
gave theoretical framework for the provision of priorities for public action. He argues that
the evaluation of the social arrangements should be made according to the level of
freedom of citizens to enhance the valuable functioning; and the main objective of this
development is the enhancement of valuable capabilities. Marshal (1890) defined the
human capital as person‟s personal wealth which included all his/her energies, qualities
and skills which had helped for his/her economic activities. Fisher (1906) said that labor
force participation in production was considered as a capital, which was helpful for
enhancement of economic growth and poverty. Solow (1956) analyzed long run growth
model. According to this model, countries use their resources efficiently and their return
diminishes as capital and labor input increase. Blaug (1972) stated that educated
person was better in production process as compared to the uneducated person
because only skilled worker had capability to increase production as he envisaged the
problems in better way and solved them.
37
3.3.1 Assessment of Poverty
To draw a line between poor and non poor is cumbersome and odd job for the
economists. So far many efforts have been made by the researchers to explain the
phenomenon of poverty and to assess its magnitude, breadth and depth. However,
there is no universal agreement among the researchers on drawing such a line of
demarcation, or the so called poverty line. As defined above, the „income poverty line‟
refers to the income sufficient enough to purchase the minimum basic needs, and this is
considered as the indicator of welfare in developed countries. But the case of
developing countries is much different where poverty is intense and distribution is more
asymmetric. In these countries, the consumption per capita is often considered to be the
preferred estimate of welfare. Some researchers regard the expenditure on
consumption per adult equivalent to be the appropriate poverty line in order to capture
differences in basic needs due to age. Other popular measures of welfare include the
calories consumption per person per day, food consumption as proportion of total
expenditure and nutritional status.
All the measures of poverty are evaluated in relation to some norms. For
example, we deem life expectancies in some countries to be low in relation to those
attained by other countries at a given date. The concept of poverty has evolved
historically and varies largely from culture to culture. Keeping in view the specific
national priorities and normative concepts of welfare, every country has its own criteria
to distinguish poor from non poor. The perception of an acceptable minimum level of
consumption tends to change as countries grow and become richer. The most popular
38
approach to measure wellbeing of people is to define a poverty line based on
expenditure function. It defines the minimum amount of resources required for achieving
a given standard of living. There are two approaches to construct poverty line. In the
first approach, poverty line is based on the household size, and it is adjusted to the
price differentials people face in different parts of the country, and also to the
demographic composition. The second approach is to draw a single poverty line for all
individuals. The per capita poverty line or level of income is then adjusted for
differences in prices and household composition. Therefore, it is pertinent to elaborate
different types of poverty lines. The common international poverty is roughly $1 per day
at 1993 purchasing power parity (PPP).
(i) Absolute Poverty line
Absolute poverty line is given in terms of standard of living. It takes into account
the minimum level of needs that are deemed necessary for survival. The poor are then
cut-off as those who do not possess this income. Here problem arises that there is no
common and concurrent definition of standard of living. All absolute poverty lines are set
in terms of goods and services. The commodity based poverty line is defined by Z;
which is a measure of the indirect utility. If utility is derived from commodities and
commodities demand depends on purchasing power, then utility level depends entirely
on this power, while assuming the prices to remain stationary:
( )U f Y or 1( )Y f U
39
This implies that for a specific level of utility, there is some income or purchasing power
or an effective expenditure level in the background (defined by Z) that is must to
achieve the specific utility level. If the quantity of goods or the level of utility so
attainable is sufficient to avoid poverty, then
1( )ZZ f U
Putting the same things in the other way round, given a poverty line that is absolute in
the space of welfare (i.e. gives Uz) there is a corresponding absolute commodity based
poverty line. Certain problems are associated with the commodity based poverty lines,
which are summarized below:
1. The referencing problem: what is the appropriate value of (Uz), the utility line.
2. The identification problem: given (Uz), what is the exact value of (Z) the
commodity value of poverty line?
For given UZ (standard of living) there is a corresponding absolute commodity based Pz.
Absolute poverty line does not vary from country to country, hence is a good tool to
compare poverty across countries.
40
(ii) Relative Poverty Line
This measure defines „poverty‟ in terms of amount tat falls below some relative
poverty threshold expenditure. According to Peter Townsend ( ), poor are poor
because their actual resources fall short of the resources held normally by other
individual in family or the society in which they live. The relative approach to define
poverty takes a moderate view. When society transforms (grows) overtime, the
perception of necessities fulfillment and hence the perception of wellbeing also
changes. Hence, if income of every household within a society increases uniformly but
the mechanism of income distribution remains the same, the perception of relative
poverty will also remain unchanged. Relative poverty line is revised upward as country
becomes well off. This form of poverty line varies from country to country.
(iii) Objective poverty line
Objective poverty line is set such that it enables an individual to achieve certain
capabilities, including health, active life and full participation in the society. Two
methods have gained popularity in recent years to develop the objective poverty line.
These are, food energy intake method, and the cost of basic needs method.
Food Energy Intake Method
According to this method, a certain level of consumption expenditure/ income is
arrived at that allows the household to obtain enough food to meet his energy
requirements.
( )k f Y which implies 1( )Y f k
41
Where, k is the level of adequate food energy intake sufficient at the margin. Now given
the minimum adequate level of calories intake mink , we have.
1
min( )Z f k
The shortcoming of this approach is that this method has no potential to concede
the relationship between food energy and income, hence suggesting the same poverty
line for urban and rural areas. However, a large number of researchers have arrived at
the conclusion that for a specific level of food intake, the poverty line in the rural areas
would be lower than in the urban areas. This is termed as the rural-urban problem.
Secondly, a rise in the relative prices of food items leads people to shift away from food
to non-food consumption. This results in poverty line to rise up. This is known as the
relative price problem.
The Cost of Basic Needs Method
The cost of basic needs method takes into account a consumption bundle that is
deemed to be adequate, comprising both food and non food components, and then
estimates the cost of this bundle for each sub group (urban rural, each region etc.). In
this approach, poverty line is calculated in the following manner.
1. The nutritional requirement for a healthy person give by 2100 to 2300 calories per
day is considered uniform for all.
2. The expenditure required to meet this food requirement is estimated. It uses a diet
that reflects the habits of household who are near to the poverty line. This food
42
component may be denoted by FZ , the calculation of which may not be easy,
particularly if diets and prices vary widely across the country.
3. Non food component ( NFZ ) is added and the basic poverty line is symbolized as:
BN F NFZ Z Z
No concordant method has yet been developed to measure the non-food component of
poverty line. However, the poverty line developed for South Korea (KIHASA 2000)
measures the cost of non-food items, as the average spending by the household in the
poorest two fifth of the income distribution.
Ravallion (1998) measured an upper poverty line by an assessment of the
income level at which the household would buy 2100 calories of food and other
necessary non-food items, and a lower poverty line by measuring the income level at
which households could just buy enough food and have no money left to buy non food
items. Given the two extremes; a household may typically buy the non-food items so
where in between. Ravallion, suggests that one might compromise on measure of the
non-food items at the midpoint of these two extreme.
This objective poverty line approach is criticized on various grounds. Due to
limitation and shortcomings, the idea of a subjective poverty line was introduced. Under
this approach, people are asked to define poverty line and the extent of poverty is then
measured by using this line. This entails different answers from different persons
according to their preferences. These can be plotted in order to get a best fit line
through econometric techniques.
43
Guarau Datt of World Bank has analyzed the Philpine data and concluded that
self rated poverty lines are high. They have climbed up rapidly overtime. The self rated
Pz (poverty line) given by poor household are slightly lower than those of the non poor
households. There is clear urban/ rural difference in the perception of the poverty line.
With the urban households setting, a money poverty line is estimated at about twice the
level of rural households. A plain explanation for this behaviour could be that inequality
prevails more in urban areas, which raises the expectations of urban people.
The choice of poverty line affects the qualification of poverty. All approaches
yield different pattern of change in poverty resource line to a great extent. For example,
in Pakistan, the policy makers do not consider a minimum standard for all citizens as
policy goal because resource constraints do not permit them act likewise.
3.3.2 Properties of Poverty Measures
Before going to discuss various measures of poverty it might be appropriate to
discuss briefly the desirable properties of a good measure. A good measure of poverty
should carry the following characteristics.
Focus: The poverty index should be independent of the income of the non poor.
Symmetry: The index should not change if any two individual exchange
incomes.
Population Invariance: Adding up or subtracting two or more identical
population should not change the poverty level. This means that poverty remains
unchanged if identical populations merge together.
44
Monotonicity: This property has three dimensions.
Poverty level must not decrease if poverty line is shifted upward.
If the proportion of poor people increases then poverty should increase.
An increase in the income of poor should decrease the poverty index.
Transfer: This property was propounded by Sen (1973). According to the
principle a regressive transfer of income between two poor people should
increase the poverty index and vice versa. The principle requires that a
progressive transfer of income between two individuals, that moves the recipient
across the poverty line, should decrease the poverty index and vice versa.
Additive Decomposability: This property implies that the poverty index should
be such that overall poverty could be related to the part of the population.
Define Limits: The poverty index should be such that it has well defined limits. In
order to make comparison easy. Generally the value of poverty index falls
between 0 and 1, where zero reflects no poverty and 1 reflects perfect poverty.
Once the poor have been distinguished from the non poor, the problem creeps up as to
how to measure poverty. Also how much poor are poor? The subsequent section of
study serves the different measures of poverty.
3.4 Measures of Poverty
Prominent economists an researchers have devised different statistics for measuring
the extent of poverty on national/regional level. Here we discuss the most important
formulations.
45
(a). Head Count Ratio (HCR)
The „Head count ratio‟ expresses the number of poor as a proportion of the
population living below the poverty line. It is denoted by Po, Mathematically,
0
1( )
p
i
NP I Y Z
N N
N= Total Population. Np= No of poor, Z is poverty line, Yi is income of household.
I(.) is an indicator function that takes on a value of 1 if the expression (Yi < Z) holds
true and zero otherwise. If Yi falls short of poverty line (Zi), then I(.) equals to 1 and
household will be reckoned as poor. The big advantage of HCR recognized by
researchers is that:
It is plain (easy) to construct;
It is simple to use; and
Its interpretation is not fuzzy as this measure has well defined limits.
However, this measure is not free from demerits. It disadvantages are:
It has nothing to do with intensity of poor or depth of poverty; this implies that
head count ratio does not translate clearly as how poor they are.
It violates the principle of transfer badly, transfer of income between two or more
households does not change the status of poverty that is, and they remain below
poverty line.
46
The estimates by HCR are not computed for households rather these are
calculated for individuals. In order to calculate the percentage of household it is
assumed that all household members maintain the same level of well being.
(b). Poverty Gap Index
Poverty gap index indicates the extent to which individuals fall below the poverty
line (by whatever way it is defined). It is denoted by P1, and expressed as under:
1
1 NGP
N Z
NG is the income gap and it measures the intensity of poverty it is given as:
( )( )N i iG Z Y Y Z
Y is income of the household, and Z is the poverty line.
Value of poverty gap lies between zero and one. A zero poverty gap implies no one is
poor and unit gap indicates that every individual requires an amount of money equal to
poverty line to get the minimum standard of living.
This measure estimates the cost of eradicating poverty, as it spells how much
income households require in order to bring them at a level above the poverty line.
Verily, this index is very helpful in evolving the poverty alleviation budget, by providing
the valuable information about minimum resources required dealing with the precarious
situation of poverty. This measure however, lacks to assimilate the effect of such
income transfer between poor where transfer does not make anyone non poor. This
47
measure potently satisfies the axiom of monotoncity and violates transfer property and
could not assign a greater weight to the income gap of poorer person.
(c) Squared Poverty Gap
The Squared poverty gap is just a weighted sum of poverty gaps, where weights
are the proportionate poverty gaps themselves. It is denoted by P2. Mathematically,
2
2
NGP
N
, where G is income gap given by:
( )( )N i iG Z Y Y Z , where Z is the poverty line.
This index solves the problem of inequality among the poor as more weight is given to
poorer amongst the poor. This measure has no upper limit making it difficult to interpret.
(d) Foster, Greer-Thorbecke Index (FGT)
Foster, Greer and Thorbecke (1984) introduce a more flexible and plausible
poverty measure. This measure is flexible and can readily be made more or less
sensitive to poverty. This index can be disaggregated for population sub groups.
Mathematically
1( 0)
a
n
a
GP where a
N N
where ( )( )n i iG Z Y Y Z and is measure of sensitivity of the index of poverty. Z is
the poverty line. HCR, PGI and SPGI are limiting cases of FGT, such that 1 , it gives
48
the poverty gap. 2 it becomes the squared poverty gap and so on. FG index can be
disaggregated for population sub groups and the contribution of each group to overall
poverty can be worked out. FGT has following distinct properties.
For all 0 the measure is strictly decreasing in the living standard of poor.
For 1 it implies the poorer one is, the increase in his measured poverty due to fall in
the standard of living will be deemed greater.
The measure is strictly convex in income and weakly convex for 1 .
FGT index satisfies the properties of monotonicity and transfer for 1 and violate the
transfer sensitivity axiom at 1 but satisfies this property for 2 . The FGT
measure is also additively decomposable with population share weight.
(e) Sen Index
A.K. Sen (1973) put forward an index which contains the effects of the number of
poor, depth of their poverty and poverty distribution with in the group symbolically. It is
given by:
0 1 1p
s pP p GZ
The definitions are as under:
P0= Head count Ratio (HCR)
P = Mean income of poor
49
PG = Gini coefficient of inequality among the poor, and 0 1PG
The values of Sen index lie between zero and one: 0 1SP , 1SP means everyone is
poor and 0SP means no one is poor. Different factors contribute individually to the
Sen index. The income gap represents poverty as measured by the proportionate gap
between the mean income of poor and the poverty line income. The Gini coefficient
measures the inequality among the poor and head count expresses the proportion of
population below poverty line.
Sen Index is considered as a comprehensive indicator of poverty, however it
cannot be decomposed into its constituent components. Therefore, it is rarely used in
practice. The Sen Index satisfies the properties of monotenity and transfer and has the
ability to give greater weight to the income gap of poorer person but it violates both
transfer and sensitivity axioms.
(f) The Sen- Shorocks-Thon Index
Another index satisfying all desirable properties is the Sen-Shorock-Thon index.
It is given as:
0 1 1SSTP P P G
This index is the product of HCR, PGI and the Gini coefficient. The Gini coefficient is
close to indicating inequality in incidence of poverty gap. It is a decomposable measure.
All measures of well-being and poverty are imperfect, but this is not an argument
to avoid measuring poverty. Instead one has to do this practice but with caution.
50
Inconsistency between two indices can be illustrated by comparison of India and
Indonesia. Both countries differ only negligibly in terms of Gini index (India=32.5,
Indonesia=34.3). However, 79.9% of the Indian population lives on below $2 a day,
while 52.4% of the population of Indonesia subsists at the same level of poverty (World
Bank, 2004a:55, 61).
(g) The Basic Needs Gap Index
Paul Streeten is considered as the pioneer of the basic needs fulfillment
approach. For the first time, he described why basic need policy, the feasibility of its
implementation, the search for a suitable yardstick to measure the role of governments
and political activists in its administration and management is crucial. Streeten (1980)
defends this approach by defining it as the one that spells out (in considerable detail)
human needs in terms of health, food, education, water, shelter, transport and non-
material needs like participation, culture, identity and a sense of purpose in life and
work, which interact with the material needs. He says, “if we judge policies by reduction
of suffering, the criterion of basic needs fulfillment scores higher than that of reducing
inequality, eliminating or reducing unemployment or alleviating poverty.”
The Basic Needs Gap Index (BNGI) is used by researchers as a proxy of basic
needs fulfillment (BNF), for instance by Hassan (1997) and Kipanga (2007). BNGI
represents the dependent variable, which is measured by the following formula:
ptntBNG
51
This measures the difference between the mean expenditure on basic needs nt
in the region and the mean income of the poor in that region, pt . The mean income of
the poor generally equals their consumption and therefore the study takes income into
consideration instead of consumption.
When the above equation is expressed as a ratio of nt , this gives the index:
ntptntBNGI ][ , which implies ntptBNGI 1
Generally the income of the poor is less than the mean expenditure on basic
needs ( pt nt ) and the index lies between zero and one. However, in rare cases, it
may happen that income of the poor exceeds the level expenditure on the basic needs
( pt nt ). In such a situation, BNGI will be less than zero. The smaller value of index
indicates better performance and higher value shows the worst condition of the region.
We postpone the construction of BNGI till chapter-4 on data and variables.
3.5 Concluding Remarks
During 1980‟s and 1990‟s the „Washington consensus‟, also known as the “neo-
liberal” policy prescriptions, remained the general advice for the developing countries in
which fiscal discipline, privatization and market liberalization were the main ingredients.
These policies were suggested in response to the real problems in Latin Americas. The
governments concerned were facing the problems of deficit, inefficiency of the public
sector enterprises, enactment of protection policies, high inflation and unemployment
etc. Obviously, there would have been definite gains if these policies could be
52
implemented in letter and spirit in the right direction. But the problem was that most of
these policies became the victim of futile exercise at the government levels rather than
serious efforts leading to more equitable and sustainable growth. The suggested
policies were supposed to be carried out in the developing countries where the markets
were more often imperfect and information system very inefficient that result into the
failure of invisible hand6.
Poverty is a complex phenomenon and it has many dimensions. In case of
Pakistan, as also in other developing countries, majority of people have low income and
they also suffer from lack of access to basic needs. Pakistan is more relying on
remittances to alleviate poverty, which permit families to maintain or enhance
expenditure on basic needs like food, clothing, housing and health. It also allows
masses to increase expenditure on durables and non durable goods, real estates and
on human capital accumulation; and thus to enhance their living standards.
The nutshell of all discussion is to answer three questions. These are:
1)- How can standard of living be increased?
2)- What is meant by minimal standard of living?
3)- Which single index or measure can be used for overall severity of poverty?
6 The free market ideology dates back to Adam Smith who argued that market forces drive the
economy to efficient outcomes. However the theory ignored the highly restrictive conditions of the
perfect competition and complete markets, which are seldom found in the developing world.
53
CHAPTER 4
DATA AND VARIABLES
For any empirical analysis, reliable and disaggregated data is crucial. Data
availability and its quality is a major hurdle in testing various economic phenomena. For
better and consistent results, one needs reliable data over a longer period of at least
thirty years. To see the basic needs gap (BNG) and its determinants across different
regions of Pakistan, the appropriate economic unit would be district. We are relying on
the data compiled by the Federal Bureau of Statistics [FBS], Government of Pakistan
under the title “Pakistan Social and Living Standards Measurement Survey (PSLM)
(Provincial/District)” formerly called the HIES. The requisite published data is available
at district level only for three years 2004-05, 2005-06 and 2006-07. Therefore we are
compelled to use the provincial level data of HIES.
4.1 Limitations of the Grouped Data
Although grouped data allows us to see changes in basic needs requirement
over longer continuous periods and to identify the pattern of changes in basic needs
gaps across different regions of Pakistan; however grouped data has certain limitations.
For instance, it displays only the mean income and share of quintiles or deciles. It leads
to the assumption that all the individuals within a group have the same income
inequality, which may not be realistic. Obviously, this may distort the estimates of
income distribution within each region.
The study uses most of the data from HIES/PSLM, which is subject to two types
of errors i.e. sampling and non-sampling errors. To overcome the problem of sampling
54
errors, FBS provides training to enumerators and other staff members. FBS also uses
latest data entry (software) programs equipped with the built-in consistency checks.
To control for the non sampling errors is somewhat difficult job in such mass level
surveys, because most of people are illiterate especially in rural areas who do not keep
any record of their income and expenditure. The goods are most often exchanged
through barter and agricultural products are not weighed properly, which leads to
incorrect estimation of the household income and expenditure. Last but not least is the
exclusion of the poorest household from the sample who are nomads and do not have
permanent dwellings.
We are considering different regions of Pakistan and the data is derived from the
same sources, so the nature of biases (if any) will remain the same, i.e. the sampling
errors will be uniformly distributed across all the regions, and therefore the validity of
results will not be affected.
4.2 Data Sources
As mentioned above, the main source of data for the present study is the
household income and expenditure survey (HIES). The survey started in July 1963 at
national level and continues to date with normal intervals of 3 years with the exception
of some gaps. In 1990-91, the document was renamed to Household Integrated
Expenditure Survey (HIES) and the questionnaires were revised according to the
needs.
HIES is not available for the year 1993-94 (due to the loss of some
questionnaires or for some technical reasons). However the required data for that year
55
used in this study is taken from “Fifty Years of Pakistan”, a publication of the Federal
Bureau of Statistics, Government of Pakistan.
The available Issues are as follows (20 surveys):1963-64, 1964-65, 1965-66,
1968-69, 1969-70, 1970-71,1971-72, 1979, 1984-85, 1985-86, 1986-87, 1987-88, 1991
(only Pakistan- R/U), 1992-93, 1996-97, 1998-99, 2001-02, 2004-05, 2005-06 and
2007-08. This data is available in two formats; the aggregate data in published form (for
all the above mentioned years) and micro data (soft form) available since 1987-88.
For the years 1963-64 to 1971-72 and 1991-92, HIES published data set is
available at national level only, whereas it is available at provincial level with rural
urban bifurcation from the year 1979 onwards. This study covers all the four provinces
(settled areas), i.e. Punjab, Sindh, KPK and Baluchistan. The federally administered
tribal areas, the mountainous northern areas and Azad Kashmir are not included in the
present study due to the non availability of data and differences in economic
characteristics from rest of the country. The four provinces account for major share in
population of the country and provide sufficient information for the purpose of this study.
Another document, the Pakistan Integrated Household Survey (PIHS) published
by the FBS, Government of Pakistan Islamabad (1995-96, 1996-97, and 1998-99)
provides data on the social indicators at the national and provincial levels. The major
indicators for health and education are taken from this survey.
In addition to the above mentioned sources, some other important documents
have also been consulted, like the Pakistan Economic Survey published by the Finance
Division, Government of Pakistan (various issues), the Demographic and Health
56
Surveys published by the National Institute of Population Studies, The Pakistan Labor
Force Survey published by the FBS, Various issues of the Fifty Years of Pakistan
published by the FBS, World Development Reports and World Development Indicators
published by the World Bank, Human Development Reports published by the United
Nations Development Program (UNDP), Data banks from International Monetary Fund
(IMF), and Social Development in Pakistan (various issues) published by the Asian
Development Bank. Different papers and publications are also consulted for data
consistency, which include Amjad. R and Kemal. A.R (1997), Cheema I.A. (2005), Irfan.
M. (2007), Jamal.H. (2006), Qureshi.S.K and Arif. G.M, (2001), Ellahi, Mahboob, Khan
S.R. Rafi (1999), Shirazi N.S (1993), Zaidi S.A (2000).
4.3 Variables Suggested for the Preliminary Model
The preliminary model is discussed in detail in the next chapter. The dependent
variable is the Basic Needs Gap Index (BNGI), which is expected to depend on the
following ten (10) explanatory variables. The model may be written in the general format
as under.
BNGI = f(YPc, SPc, Rem, HS, HE, DR, Un, B20, T2B, HCI)
The intuitive list of explanatory variables is given below. However, it is possible that
some of these variables may be insignificant, which will become clear after conducting
the specification tests.
1. Ypc = average monthly per capita income
2. Spc = average monthly per capita savings
3. Rem. = remittances (domestic and foreign % of monthly income)
4. HS = household size
57
5. HE = higher education (BA/BSc both sexes)
6. DR= dependency ratio
7. Un. = labor force unemployment rate
8. B20= share of income held by bottom 20%
9. T2B= ratio of income of top 20% to bottom 20 %
10. HCI = human capital index
As discussed above, we have derived the relevant data from various issues of
HIES/PSLM. According to HIES (2007-08), household is either a single person or group
of persons, who live together and share cooking and other essentials of living. In the
multi-person household, the individuals may pool the whole amount of income or a part
for general consumption. The main feature is that they reside together and don‟t have
alternative residence. The main sampling unit in all surveys is the household.
Detailed discussion of the concerned data and justification of these explanatory
variables is given below. The construction of dependent variable, being very complex, is
postponed till the next section.
4.3.1 Per Capita Income (Monthly in Rs) [Ypc]
Per capita income is a powerful indicator representing economic growth of a
country. The per capita income is obtained by dividing national income (usually GDP) by
population of a country. It roughly indicates how fast an economy is flourishing. In
contrast to the gross GDP, the per capita income also covers the demographic aspect
of economy and so it is considered to be more comprehensive indicator.
The role of national per capita income in reduction of poverty is one of the most
discussed topics. Many studies found strong statistical association between per capita
income and poverty indicators. According to World Bank (1999), relationship between
58
infant mortality rate and GNP per capita, ratio of female to male literacy1 and per capita
income, and between average consumption and the incidence of income poverty is
negative. This indicates that as per capita income rises, the national poverty indicators
fall. The present study includes per capita income as one of the potential regressors for
the preliminary model.
Per Capita Income for country level as well as for the four provinces with rural
urban bifurcation is shown in Table 4.1. A glance at the data reveals that per capita
income for the economy increases over the study period. This gives the evidence that
GDP grew at higher rates than population. Interestingly, rural per capita income grew
more moderately as compared to the urban per capita income. Per capita income trends
for all the four provinces synchronized with the trend shown by economy as a whole.
However, the per capita income depicts a downward trend after 2005-06.
4.3.2 Per Capita Saving (Monthly in Rs) [Spc]:
Saving is regarded as basis for capital formation. Countries which succeeded in
setting aside a reasonable fraction of income for saving and investment have managed
to develop miraculously. For achieving a high average growth in real GDP, the saving
rate needs to be in line with the investment requirements. Since investment leads to
capital formation and an increase in productivity, then the per capita saving has
conceptually positive impact on poverty reduction. High rates of saving and investment,
and rising productivity are the foundation for rapid and sustainable growth, which
indirectly improve the living standard of the inhabitants. We included this variable in the
1 (non-income measure of poverty)
59
tentative list of regressors, although there is a difference of opinion among economists
about the causal relation of saving and growth.
Table 4.2 presents data on per capita saving for Pakistan and for the four
provinces with rural-urban break-up. It is apparent from the Table that PC monthly
saving increased by 125% for Pakistan as a whole, by 341% and 153% for rural and
urban Pakistan respectively. This significant increase in saving and obviously a
substantial increase in income of people living in rural area of Pakistan might be due to
their immigration to foreign countries in search of employment and their remittances
sent home. This led to a significant increase in their propensity to save. Same increase
is traceable for rural Punjab and rural Sindh, which is 1000% high in 2007-08 as
compared to 1979-90. On the other hand, per capita saving fell drastically in KPK and
Balochistan except for rural KPK and urban Balochistan, where marginal increases in
per capita saving is substantial.
4.3.3 Remittances (domestic and foreign as % of monthly income) [Rem]:
Remittances are very important source of income for poor and developing
countries. In some cases it is more than 20% of GDP (Global Economic Prospects
2011). As the remittances are directly received by the households, they are expected to
reduce poverty directly. The impact of remittances on poverty reduction can be
understood from both micro and macro perspectives. However to capture this impact,
there is no formal framework. Chimhowu et al, (2005). Uruci and Gedeshi (2003) found
that about 70 percent of international legal migrants send their money in order to
support the essential needs of family.
60
The present study also used remittances (Rem) as a determinant of BNGI.
Remittances are sum of both foreign remittances (F) and domestic remittances (D).
Data on foreign remittances (F) and domestic remittances (D) is not available separately
before 1990. This data is available in combined form under the column “gifts and
assistance” (GAs) for 1980s figures. This ratio of “foreign remittances” to “gifts and
assistance” (F/GAs) is taken for 1990s onwards and this ratio is multiplied by the 1980s
figures to get foreign remittances out of “Gifts and Assistance”. In the same way, values
of D are obtained and sum of the two (i.e. F and D) gives us estimates of remittances
(Rem). However, separate data for gifts and assistance, foreign remittances and
domestic remittances is available in the HIES from 1992-93 onwards.
In our preliminary model, we use the combined foreign and domestic remittance
as one of the regressor. Table 4.3 (a) presents foreign remittances (F) as percentage of
total monthly income. This percentage witnesses no increase up to 1987-88. It suddenly
increased in 1990s and kept on increasing at steady rate up to 2005-06 but it shows a
downward trend thereafter. Only for Balochistan, the percentage of total monthly income
by foreign remittances reflects an increase beyond 2005-06.
Table 4.3 (b) presents percentage of total monthly income by domestic (D)
remittances. It is apparent that during 1980s no considerable change is seen. However,
from the very beginning of the 1990s, it jumped to 4.1% for overall Pakistan and to
5.53% and 1.76% respectively for rural and urban areas. The change in percentage of
total monthly income by domestic remittances exhibits same pattern in Punjab and KPK.
However, Sindh and Balochistan depict a different story. In these provinces the change
is steady rather than by strides.
61
Table 4.3 (c) reports the percentage of monthly income by total remittances
(F+D). It is obvious that no significant change is observable till 1987-88. However, an
unbelievable change is witnessed from 0.65% in 1987-88 to 7.01% in 1992-93. This
sudden increase in Rem can be attributed to immigration of labor force to Middle East. It
oscillated down for the spell 1993-94 and 1996-97, but again it gained momentum and
showed consistent rise.
62
Table 4.1 Per Capita Income (Monthly) (Rs) CPI 2005=100
PAKISTAN PUNJAB SINDH KPK BALOCHISTAN
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 1128 929 1402 1062 935 1305 1230 841 1514 1215 1023 1599 1079 876 1292
1984-85 1242 1105 1563 1182 1079 1430 1372 1054 1714 1331 1231 1929 1201 1171 1484
1985-86 1241 1106 1562 1210 1110 1748 1358 1056 1693 1157 1113 1424 1274 1208 1622
1986-87 1277 1123 1613 1221 1084 1548 1438 1149 1760 1215 1174 1429 1377 1356 1485
1987-88 1253 1091 1644 1217 1088 1576 1384 1011 1817 1173 1137 1343 1301 1250 1593
1992-93 1305 1133 1738 1297 1175 1646 1408 1020 1898 987 923 1456 1156 1118 1417
1993-94 1287 1073 1776 1275 1131 1629 1388 919 2080 1041 964 1405 1123 1059 1467
1996-97 1387 1271 1650 1455 1385 1626 1514 1241 1772 1039 996 1308 1175 1097 1464
1998-99 1337 1110 1882 1335 1122 1863 1482 1100 2338 1064 964 1639 1434 1409 1706
2001-02 1241 1038 1737 1261 1095 1673 1290 905 1892 1073 995 1533 1217 1133 1615
2004-05 1435 1166 2017 1449 1211 1965 1552 1092 2171 1219 1107 1786 1286 1175 1715
2005-06 1671 1460 2084 1765 1612 2085 1719 1292 2157 1428 1303 2098 1091 1018 1322
2007-08 1569 1342 2034 1648 1502 1965 1628 1067 2269 1315 1226 1768 1048 881 1479
Source: HIES/PSLM (various issues)
63
Table 4.2 Per Capita Savings (Monthly) (Rs) CPI 2005=100
PAKISTAN PUNJAB SINDH KPK BALOCHISTAN
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 88 36 164 68 45 112 96 8 160 149 24 397 110 39 186
1984-85 85 65 131 78 60 110 64 26 105 153 102 448 114 111 142
1985-86 78 71 95 84 83 101 65 38 95 46 32 128 176 167 222
1986-87 54 41 81 44 32 45 55 28 81 46 42 68 214 217 201
1987-88 60 34 122 52 35 100 79 10 158 36 25 90 147 135 219
1992-93 38 -7 256 60 26 154 25 -71 148 -62 -85 111 99 87 182
1993-94 28 5 82 51 31 100 -15 -63 47 -3 -15 61 43 22 199
1996-97 143 176 67 201 244 92 87 150 26 18 15 42 104 86 172
1998-99 72 46 136 102 74 171 64 43 107 -44 -65 83 111 103 174
2001-02 79 46 159 113 73 212 27 12 50 18 -17 217 130 93 306
2004-05 84 32 195 96 37 225 60 8 129 87 49 279 49 16 175
2005-06 221 265 135 240 328 55 250 306 193 149 89 474 81 72 109
2007-08 195 159 268 225 233 207 231 86 396 54 33 161 66 4 225
Source: HIES/PSLM (various issues)
64
Table 4.3 (a) Percentage of Total Monthly Income by Foreign (F) Remittances
PAKISTAN PUNJAB SINDH KPK Balochistan
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 0.30 0.31 0.38 0.39 0.30 0.58 0.14 0.03 0.19 0.47 0.63 0.25 0.12 0.22 0.07
1984-85 0.39 0.29 0.67 0.37 0.29 0.52 0.59 0.10 1.08 0.28 0.27 0.34 0.54 0.56 0.31
1985-86 0.33 0.23 0.61 0.34 0.24 0.61 0.40 0.06 0.74 0.30 0.29 0.35 0.25 0.22 0.44
1986-87 0.37 0.29 0.60 0.41 0.33 0.57 0.40 0.06 0.76 0.27 0.25 0.39 0.27 0.31 0.09
1987-88 0.30 0.21 0.54 0.33 0.26 0.49 0.37 0.03 0.75 0.16 0.16 0.21 0.14 0.14 0.14
1992-93 2.91 2.65 3.34 2.96 2.5 3.85 1.78 0.47 2.68 5.8 6.27 3.66 1.22 1.25 1.05
1996-97 2.15 1.63 3.04 2.25 1.43 3.98 1.12 0.33 1.64 3.54 3.49 3.79 2.04 2.03 2.06
1998-99 3.24 3.15 3.37 3.38 2.78 4.26 0.93 0.14 1.5 9.06 9.44 7.74 1.29 1.45 0.35
2001-02 3.12 3.13 3.12 3.36 3.04 3.87 0.87 0 1.58 7.33 7.52 6.6 1.97 2.17 1.29
2004-05 3.58 3.45 3.75 3.84 3.08 4.87 1.37 0.76 1.79 8.02 8.57 6.3 2.04 2.15 1.76
2005-06 4.42 5.08 3.51 5.13 5.09 5.19 0.72 0.33 0.96 9.42 10.71 5.12 1.56 1.61 1.43
2007-08 4.31 5.48 2.74 4.83 5.35 3.95 0.44 0.06 0.65 10.48 11.91 5.45 1.74 1.8 1.66
Source: HIES/PSLM and Labour Force Survey (various issues)
65
Table 4.3 (b) Percentage of Total Monthly Income by Domestic (D) Remittances
PAKISTAN PUNJAB SINDH KPK Balochistan
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 0.35 0.44 0.31 0.48 0.48 0.48 0.13 0.09 0.13 0.53 0.75 0.22 0.09 0.18 0.04
1984-85 0.46 0.41 0.55 0.46 0.47 0.43 0.57 0.28 0.76 0.32 0.32 0.30 0.41 0.44 0.19
1985-86 0.40 0.33 0.50 0.43 0.39 0.51 0.38 0.18 0.52 0.34 0.35 0.31 0.19 0.17 0.28
1986-87 0.44 0.41 0.48 0.51 0.54 0.47 0.38 0.17 0.53 0.30 0.29 0.35 0.21 0.24 0.06
1987-88 0.35 0.30 0.44 0.41 0.41 0.40 0.36 0.09 0.53 0.18 0.18 0.19 0.11 0.11 0.09
1992-93 4.1 5.53 1.76 4.87 6.12 2.43 0.55 0.67 0.47 9.23 10.13 5.14 -0.09 0.11 -1.18
1996-97 3.04 4 1.36 3.39 4.17 1.73 0.37 0.08 0.56 6.84 7.53 3.57 0.03 0.06 -0.07
1998-99 3.67 5.32 1.32 4.4 6 2.02 0.09 0.33 -0.08 10.4 12.27 3.99 -0.08 -0.09 -
2001-02 3.97 5.43 1.83 4.59 5.69 2.82 0.13 0.13 0.16 10.71 12.61 3.57 0.47 0.27 1.15
2004-05 4.29 5.74 2.49 5.19 6.61 3.3 0.96 0.9 1 8.76 10.01 4.85 1.21 1.38 0.76
2005-06 3.84 5.34 1.75 4.61 5.64 2.94 0.07 0.13 0.03 8.69 10.57 2.45 0.8 0.96 0.42
2007-08 4.05 5.65 1.89 4.86 6.14 2.73 0.38 0.13 0.51 8.55 9.94 3.67 0.34 0.28 0.42
Source: HIES/PSLM and Labour Force Survey (various issues)
66
Table 4.3 (c) Percentage of Monthly Income by Remittances (F+D)
PAKISTAN PUNJAB SINDH KPK Balochistan
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 0.65 0.74 0.69 0.87 0.79 1.06 0.27 0.12 0.32 0.99 1.38 0.47 0.20 0.40 0.11
1984-85 0.85 0.70 1.22 0.82 0.76 0.95 1.16 0.37 1.84 0.60 0.59 0.65 0.95 0.99 0.50
1985-86 0.73 0.57 1.11 0.77 0.63 1.11 0.78 0.25 1.26 0.64 0.64 0.66 0.45 0.39 0.73
1986-87 0.81 0.70 1.08 0.93 0.87 1.05 0.78 0.23 1.29 0.57 0.54 0.74 0.48 0.55 0.15
1987-88 0.65 0.51 0.97 0.74 0.66 0.89 0.72 0.12 1.28 0.35 0.34 0.40 0.25 0.26 0.23
1992-93 7.01 8.18 5.10 7.83 8.62 6.28 2.33 1.14 3.15 15.03 16.4 8.8 1.13 1.36 -0.13
1993-94 7.18 8.29 5.59 7.84 8.22 7.18 2.9 2.23 3.29 15.37 17.16 8.23 2.5 2.8 0.9
1996-97 5.19 5.63 4.40 5.64 5.60 5.71 1.49 0.41 2.2 10.38 11.02 7.36 2.07 2.09 1.99
1998-99 6.91 8.47 4.69 7.78 8.78 6.28 1.02 0.47 1.42 19.46 21.71 11.73 1.21 1.36 0.35
2001-02 7.09 8.56 4.95 7.95 8.73 6.69 1 0.13 1.74 18.04 20.13 10.17 2.44 2.44 2.44
2004-05 7.87 9.19 6.24 9.03 9.69 8.17 2.33 1.66 2.79 16.78 18.58 11.15 3.25 3.53 2.52
2005-06 8.26 10.42 5.26 9.74 10.73 8.13 0.79 0.46 0.99 18.11 21.28 7.57 2.36 2.57 1.85
2007-08 8.36 11.13 4.63 9.69 11.49 6.68 0.82 0.19 1.16 19.03 21.85 9.12 2.08 2.08 2.08
Source: HIES/PSLM and Labour Force Survey (various issues)
67
4.3.4 Household Size (HS)
As poverty is multi dimensional and complex phenomenon, various Economic,
social, political and demographic factors influence it. Household size is intimately
connected with poverty status. It is observed that in countries like Pakistan the extended
family has diverse effects on poverty. In large family, the young members continue to
live with parents even after adulthood and marriages. They contribute their earning to
common pool of family. In the extended family, the young and old aged have full
insurance and security by living in combined family. They derive their livelihood from
common pool during their unemployment spell. This hinders mobility both in space and
in occupation. Due to large proportion of dependents and least participation of women in
income generating activities, the poverty persists in joint families. The system of
inheritance, which divides land and other assets among the heirs, also affects poverty to
some extent. Large family having more dependents and less earning hands is certainty
prone to poverty. Conceptually the household size is positively related with poverty.
Table (4.4) presents households size for Pakistan and the four provinces with
rural urban bifurcation. For Pakistan, the household size at the outset of study period is
6.1 and household size increases marginally up to 6.58 toward the end of the study
period. It is worth mentioning that household size remained higher in rural Pakistan as
compared to urban Pakistan till 1996-97. However, a reverse pattern is visible after this
period till the end of study period and almost same pattern is observed in Sindh and
KPK. The household size remained low in rural areas of Balochistan as compared to
urban regions throughout the study period.
68
4.3.5 Higher Education (BA/BSc both sexes) (HE)
Higher literacy rate and primary education is argued to be important for economic
growth and development. In many developing countries, secondary and higher
education is neglected due to emphasis on primary education by donor agencies.
According to Mamphela Ramphele, managing director for human development,
World Bank (2002), “There is no way we can succeed in the eradication of poverty if the
developing world is not part of knowledge creation, its dissemination and utilization to
promote innovation. Higher education is a critical factor in making this possible and
must be part of any development strategy.”
When David Lavin, the Grawemeyer Award winner in education 2009, was asked
that “how can we break the generational cycle of poverty?” he answered that you make
sure that the disadvantaged, especially women, have access to a college education.
By higher educaton we mean the percentage of earners out of the total labor
force, having degree level (BA/BSc) education. It also implies that degree holder with no
employment are excluded since they have no impact on the dependent variable. Data
for the higher education (Having BA/BSc degree) is obtained from HIES. Data for the
year 1993-94 is not available and is filled with five year moving average. Table (4.5)
demonstrates the percentage distribution of earners (both sexes) by degree level
education. A glance at the table asserts that distribution was more skewed towards
urban Pakistan in 1979. Distribution is awfully low in rural Pakistan. In all provinces this
percentage increases at rapid rates as in urban areas as compared to rural area. This
however is not surprising because infrastructure lacks in rural Pakistan. As a result of
which migration at mass level is taking place from rural to urban area.
69
Table 4.4 Household Size
PAKISTAN PUNJAB SINDH KPK BALOCHISTAN
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 6.1 6 6.4 6 5.9 6.2 6.4 6.3 6.5 6.3 6.1 6.7 6.3 5.8 7
1984-85 6.21 6.05 6.65 6.28 6.15 6.64 6.19 5.85 6.59 6.37 6.25 7.11 5.33 5.22 6.44
1985-86 6.34 6.17 6.77 6.2 6.07 5.56 6.66 6.31 7.1 6.68 6.65 6.83 5.71 5.59 6.47
1986-87 6.46 6.32 6.79 6.31 6.19 6.63 6.71 6.49 6.98 6.82 6.77 7.13 6.22 6.09 6.92
1987-88 6.3 6.16 6.66 6.18 6.06 6.52 6.41 6.15 6.73 6.83 6.75 7.31 5.9 5.76 6.9
1992-93 6.4 6.3 6.66 6.55 6.44 6.78 6.42 6.19 6.73 7.1 7.23 6.26 5.83 5.66 7.16
1993-94 6.21 6.18 6.4 6.29 6.16 6.664 6.21 6.11 6.02 7.02 7.15 6.86 5.62 5.56 6.96
1996-97 6.21 6.14 6.37 6.13 6.04 6.38 5.87 5.48 6.29 7.11 7.18 6.72 5.85 5.71 6.46
1998-99 6.77 6.82 6.65 6.5 6.48 6.54 6.74 6.87 5.57 7.8 7.84 7.63 7.5 7.43 7.5
2001-02 6.96 7 6.87 6.54 6.5 6.63 7.54 7.87 7.08 7.66 7.67 7.55 7.63 7.56 7.96
2004-05 6.75 6.8 6.63 6.55 6.56 6.54 6.71 6.84 6.54 7.71 7.69 7.77 6.88 6.79 7.27
2005-06 6.83 6.93 6.65 6.46 6.43 6.53 7.02 7.57 6.53 7.96 8.03 7.61 7.51 7.28 8.4
2007-08 6.58 6.72 6.31 6.33 6.35 6.28 6.5 6.97 6.04 7.63 7.71 7.23 7.75 7.59 8.17
Source: HIES/PSLM (various issues)
70
Table 4.5 Percentage Distribution of Earners (Both Sexes) by Degree Level Edu:
PAKISTAN PUNJAB SINDH KPK BALOCHISTAN
YEARS Overall Rural Urban YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 0.72 0.13 1.6 1979 0.52 0.11 1.31 1.46 0.2 2.39 0.45 0.09 1.13 0.52 0.24 0.83
1984-85 1.83 0.7 4.68 1984-85 1.68 0.53 2.74 3.64 0.43 7.75 0.98 0.45 4.06 4.37 3.95 8.02
1985-86 1.82 0.55 5.1 1985-86 1.37 0.44 4.02 3.36 0.68 7.07 1.1 0.8 2.91 1.11 0.55 4.37
1986-87 2.03 0.48 5.76 1986-87 1.47 0.3 4.55 4.02 0.89 8.27 1.32 0.74 4.2 0.79 0.47 2.46
1987-88 2.09 0.59 5.87 1987-88 1.57 0.54 4.54 3.9 0.46 8.53 1.29 0.84 3.34 1.71 1.15 5.17
1990-91 2.05 1.03 4.46 1992-93 1.6 0.75 4.25 4.52 1.35 8.88 1.74 1.1 6.48 0.96 0.39 5.51
1992-93 2.24 0.89 6.05 1993-94 1.69 0.62 4.75 4.98 1.36 9.52 1.52 0.96 4.87 1.2 0.62 4.4
1993-94 2.38 0.9 6.1 1996-97 2.14 0.92 5.68 7.51 2.77 12.4 1.76 1.17 5.46 1.36 0.46 5.24
1996-97 3.16 1.22 8.02 1998-99 1.85 0.74 5.04 5.23 2.14 10.13 2.28 1.79 5.1 2.78 2.36 5.81
1998-99 2.8 1.27 6.89 2001-02 2.2 0.91 5.85 4.81 1.91 10.36 2.78 2.02 7.11 3.17 2.33 7.72
2001-02 3.02 1.36 7.55 2004-05 3.86 2.16 8.14 9.34 3.94 17.53 4.59 3.47 10.42 3.61 2.03 11.88
2004-05 5.34 2.71 11.63 2005-06 2.89 1.1 7.19 6.73 2.11 12.34 3.95 3.18 8.26 3.45 2.26 8.24
2005-06 4.03 1.67 9.18 2007-08 3.43 1.71 7.55 6.93 2.7 12.47 4.24 3.26 9.49 3.84 1.48 10.43
2007-08 4.41 2.11 9.45
Source: HIES/PSLM (various issues)
71
4.3.6 Dependency Ratio (DR)
Dependency ratio relates the number of children (0—14 years) and elders (65
years and above) to the working age population (15—64 years), expressed as
percentage. Dependency ratio has different implications; for instance, high ratio is a
sign of aging population, overburdened pension, social security requirements to the
older and non working population. The ratio is given below:
0 14 65100
15 64
number of people aged and those aged and aboveDependency Ratio
number of people aged
This can be further disaggregated as:
0 14100
15 64
number of people agedChild Dependency Ratio
number of people aged
65100
15 64
number of people aged and aboveAged Dependency Ratio
number of people aged
Larger number of dependents in the form of children and elderly members is
equivalent to smaller number of income earners or a smaller per capita income in the
household. This ratio allows quantify the burden on members of the labor force. One
might expect that a high dependency ratio would be correlated positively with the level
of household poverty. This ratio is considered a preferred social and demographic
indicator as compared to the household size since the later ignores the number of
earners. For example, large sized households with many earners may suffer low per
capita income whereas small sized households with only one significant earner may
enjoy relatively high per capita income. However, there is a difference of opinion among
the researchers regarding the relationship between the dependency ratio and the level
of poverty or inequality in distribution.
72
Idrees (2006) could found no relationship between dependency ratio and
inequality. He is of the view that, changes in the dependency ratio do not translate into
corresponding changes in incomes per adult-equivalent. Imran Sharif Chaudhry,
Shahnawaz Malik and Abo ul Hassan (2009) are of the view that dependency ratio has
a significant impact on a household‟s well being. The results support the hypothesis that
poverty will be more severe among the households with higher dependency ratio.
According to Njimanted, Godfrey Forgha (August 2006), the dependency ratio is
negatively related to poverty. The study shows that 10 percent increase in these
variables will result into 0.7164 percent fall in poverty.
Dependency ratio for Pakistan and the four provinces with rural urban bifurcation
is given in Table 4.6, which shows a mixed trend. Rural dependency ratio is much
higher than the urban. This fact is also observed by Idrees (2006). The obvious reason
is that rural Pakistan draws livelihood mainly from agriculture, which is characterized by
open and disguised unemployment. The segment of population between the age 1-14
and over 65 is more densely distributed in rural Pakistan than the urban regions. One of
the main reasons of prevalence of higher poverty in rural areas is the higher
dependency ratio. Dependency ratio was very high during 1979 in Pakistan as well as in
provinces except in case of Balochistan, where the ratio in rural Punjab and urban KPK
stood at 100. For the Pakistan as a whole, there is about 15% decline in dependency
ratio from 1979-80 to 2007-08. The ratio declined more drastically in Punjab by about
22% during the same period. A rise in the ratio in case of rural Sindh and rural
Balochistan is serious feature. We may conclude that dependency ratio is not falling
with a rate sufficient enough to make a dent in poverty.
73
Table 4.6 Dependency Ratio
PAKISTAN PUNJAB SINDH KPK BALOCHISTAN
Year Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 97 99 93 99 100 96 93 97 87 112 115 100 90 91 86
1984-85 98 101 92 93.48 95.61 88.12 92 98 84 102.3 105.3 86.48 94 95 92
1985-86 95.3 98 90 98 101 94 96 103 91 105 110 99 106.3 109.2 92.1
1986-87 100 103 94 95 98 89 95 103 87 109 113 90 107 109 105
1987-88 98 102 89 93 97 85 95 105 86 109 114 86 113 115 103
1992-93 98 104 86 96 100 86 94 105 86 110 115 86 108 110 100
1993-94 96 102 84 94 99 84 92 103 83 108 113 85 106 107 100
1996-97 97 101 90 95 97 91 93 99 87 110 114 91 96 108 107
1998-99 76 96 86 86.9 92.3 76.7 88 100 79 85.6 88.6 81.5 108 108 109
2001-02 86 92 75 82 91 73 88 98 77 96 105 81 102 106 96
2004-05 84 92 71 82 88 72 84 97 72 93 96 79 98 100 92
2005-06 81 89 69 81 88 69 84 98 71 92 94 80 97 99 93
2007-08 82 92 72 77 78 67 83 99 74 92 95 78 92 95 83
Source: HIES/PSLM and Labour Force Surveys (various issues)
74
Table 4.7 Labour Force Unemployment Rate (Un)
PAKISTAN PUNJAB SINDH KPK BALOCHISTAN
Year Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 3.0 3.6 5.2 3.5 4.3 6.2 1.8 1.8 4.0 3.3 3.5 4.6 1.5 2.2 2.2
1984-85 4.0 3.7 5.7 4.5 4.3 6.7 2.7 2.5 4.2 4.2 3.9 6.1 1.6 1.5 4.1
1985-86 3.7 3.3 4.0 4.2 3.5 4.9 2.4 1.6 2.2 3.9 3.8 4.2 1.6 1.4 2.0
1986-87 4.7 4.2 5.2 5.4 4.6 6.4 3.0 2.0 2.7 4.8 4.7 5.1 1.5 1.4 1.9
1987-88 6.0 5.4 7.6 7.2 6.4 9.5 3.5 2.1 4.9 5.9 5.7 7.1 1.4 1.1 3.5
1992-93 4.5 4.1 5.6 5.3 4.7 7.3 2.3 1.7 3.1 5.2 5.2 5.3 2.2 2.1 2.5
1993-94 4.5 4.0 5.8 5.5 4.6 8.1 2.4 2.0 3.0 7.1 6.9 8.3 2.8 3.0 2.1
1996-97 5.7 5.3 6.7 6.2 4.7 9.8 2.8 2.1 3.5 8.8 8.7 9.3 2.5 2.1 4.4
1998-99 5.6 4.7 7.6 7.7 6.2 11.7 2.8 2.1 3.7 11.7 11.7 11.8 5.9 5.7 7.0
2001-02 7.8 7.2 9.1 7.9 7.0 10.0 5.1 3.2 6.9 13.1 12.7 15.0 6.3 6.3 6.2
2004-05 7.3 6.4 9.2 5.8 4.8 8.3 4.3 2.9 5.7 12.0 12.1 11.6 3.1 2.4 5.6
2005-06 6.1 5.4 7.5 5.1 4.3 7.0 3.4 2.3 4.5 9.6 9.2 11.9 2.6 2.2 3.9
2007-08 5.0 4.5 6.0 5.1 4.5 6.6 3.1 2.0 4.3 8.9 8.7 10.0 2.7 2.1 4.8
Source: Labour Force Survey and Social Development in Pakistan (various issues)
75
4.3.7 Labor Force Unemployment Rate (Un)
Unemployment is considered one of the biggest obstacles to rapid economic
growth and poverty reduction. High unemployment can be attributed to high population
growth and aimless educational system. Population is growing at an alarming rate in
Pakistan and thus responsible for adding more people to the pool of unemployed and
poverty. It has various dimensions; on the one side unemployment-to-population ratio
and labor force participation rates are dismal and on the other hand, the female to male
unemployment ratio is crucial. Unemployment rates are higher in urban areas as
compared to rural areas.
Data for the labor force unemployment rate is obtained from Social Development
in Pakistan (various issues) and is presented in Table 4.7. Unemployment in 1979 was
not so much alarming as is evident by the figures 3.0 for overall Pakistan, 3.6 for rural
areas and 5.2 for urban areas. There has been massive influx of people in the labor
force over the past thirty years, for which our economy has no potential for absorption.
That is why the labor force unemployment rate grew by a figure of 66 percentage points
from 1979 to 2007-08. This increase is 25 percentage points for rural areas whereas it
is 16 percentage points for urban areas. Similarly labor force unemployment rate for
rural Punjab grew slowly by an average of 4.6% as compared to 5.5% for overall Punjab
and 6.45% for urban Punjab. The values for the year 1985-86 and 1986-87 are
interpolated using three years moving average. For overall Pakistan and all the
provinces in general, the labor force unemployment rate remained high in urban areas
as compared to rural areas. This rate remained high in KPK for most of the time,
whereas the rural Sindh and rural Balochistan enjoyed a lower rate during the study
76
period. A sharp increase in the unemployment rate is observed in almost all regions of
Pakistan after the year 2000. This increase might be due to many factors including the
9/11 incident. Pakistan remained the front line ally of the US and allied forces against
war on terror in Afghanistan and own tribal areas. However, in the 2007-08, all the
regions of Pakistan witnessed downward trend in labor force unemployment rate.
4.3.8 Share of Income held by Bottom 20% Population (B20)
Distribution of income is closely related with economic growth and poverty. Some
researchers like Lewis (1954) were of the view that inequality promotes saving,
investment and growth, which in turn tends to reduce poverty. However, the empirical
evidence revealed that inequality has a negative impact on development - see for
instance Alesina (1993) and Roberto Perotti (1996)
There are several ways to measure inequality. Traditionally, the size distribution
is measured in terms of relative amounts received by 10% or 20% of income earners.
The present study incorporates the share of income held by the bottom 20% of the
households (B20) as one of the regressor for its significance towards poverty. Table 4.8
presents this picture. For the year 1979-80, this share was 12.3% for overall Pakistan
(14.1% for rural areas and 10.6% for urban areas). The share however, reduced by the
year 2007-08 to 7.9% for overall Pakistan (10.5% for rural areas and 3.0% for urban
areas). Same trend prevails for urban areas of all the provinces. However, the said
share in overall and rural Balochistan has increased sharply during this period.
Likewise, rural Sindh has shown slight improvement in the indicator concerned over the
data period but with some fluctuations.
77
4.3.9 Ratio of Income of Top 20% to Bottom 20 % (T2B)
Equity and fair distribution provides a great motive for participation in a
development process. The individuals will pursue a life of their choice without any
feelings of deprivation. According to World Development Report (2006), “Institutions and
policies that promote a level playing field, where all members of society have similar
chance to become socially active, politically influential and economically productive;
contribute to sustainable growth and development. Greater equity is thus doubly good
for poverty reduction.”
A frequently used measure of inequality is the ratio of the share of income
received by the 1st quintile (top/richest 20%) to the share of income received by the 5th
Quintile (bottom/poorest 20%). The intuitive list of regressors includes this ratio as
explanatory variable and it is expected to be positively related with BNGI. These results
are similar to Kipanga (2007), where it was observed that group averages varied
positively with the levels of BINGI performances.
Table 4.9 depicts the ratio of income of top 20% to bottom 20% for Pakistan and
all provinces. This measure gives a clearer picture of income inequality. The data shows
some interesting features. Almost 90% increase in the ratio (from 3.1 in 1979-80 to 5.9
in 2007-08) is visible for Pakistan. Increase in this measure for rural Pakistan is a
somewhat moderate, but there is unprecedented increase in this ratio for urban
Pakistan (from 4.3 to 22.5) and urban Sindh (3.8 to 57.2) during the study period. The
ratio increased steadily up to 2007-08 with spikes in some years. In contrast, the rural
Sindh and rural Balochistan have shown a clear downward trend.
78
TABLE 4.8 Share of Income held by bottom 20 Percent
Pakistan Punjab Sindh KPK Balochistan
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 12.2 14.1 10.6 12.1 13.5 10.6 12.7 17.2 12.0 11.7 13.7 8.7 13.3 16.0 12.0
1984-85 9.7 12.4 10.8 10.3 12.4 10.3 11.0 15.3 13.8 8.1 11.2 8.9 8.7 11.6 13.0
1985-86 12.0 13.2 10.4 12.0 12.9 10.7 11.6 14.5 11.1 13.0 13.4 10.8 12.8 13.3 12.2
1986-87 11.8 13.2 10.1 12.1 13.4 10.1 12.3 14.7 11.7 11.5 11.7 11.2 15.5 15.6 16.9
1987-88 12.0 13.4 12.0 11.9 13.0 11.4 13.2 17.3 12.6 13.5 12.4 12.7 15.3 15.8 14.5
1992-93 11.7 12.8 9.7 11.4 12.3 9.7 10.4 13.5 9.5 13.4 14.1 10.7 13.6 13.0 12.2
1993-94 9.4 11.5 7.5 11.3 10.9 9.5 11.7 14.0 9.9 10.7 12.0 7.9 11.4 13.0 10.3
1996-97 10.6 11.3 10.9 10.8 10.7 10.8 11.4 12.8 10.9 12.1 12.4 11.4 10.8 14.8 13.0
1998-99 7.9 9.9 8.2 8.4 10.2 8.0 9.4 10.1 9.4 8.1 10.3 8.8 10.2 11.8 11.6
2001-02 9.7 13.0 4.8 8.8 10.6 5.9 9.5 18.1 3.0 14.0 16.3 5.2 10.2 11.2 7.1
2004-05 9.3 13.8 3.7 9.9 14.0 4.5 6.5 13.0 2.1 12.9 15.1 5.8 8.8 10.2 5.3
2005-06 7.6 10.4 2.2 5.9 7.7 2.0 10.3 18.9 1.9 10.2 11.9 3.4 17.2 20.2 6.4
2007-08 7.9 10.5 3.0 7.5 9.1 4.0 8.6 19.2 1.2 6.9 7.9 2.6 17.4 24.3 5.4
Source: Own calculations till 1998-99 using data from HIES/PSLM (various issues)
79
TABLE 4.9 Ratio of Income of Top 20 % to Bottom 20 %
Pakistan Punjab Sindh KPK Balochistan
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 3.1 2.1 4.3 3.0 2.3 4.0 3.2 1.4 3.8 3.6 2.4 4.9 2.6 1.6 3.3
1984-85 4.5 2.7 4.2 4.0 2.7 4.2 3.9 1.9 3.2 6.4 3.4 6.2 4.6 2.7 2.7
1985-86 3.2 2.5 4.4 3.1 2.6 4.2 3.5 2.0 4.1 2.8 2.5 4.2 2.8 2.5 3.4
1986-87 3.3 2.5 4.6 3.1 2.4 4.7 3.4 2.1 3.9 3.3 3.1 4.0 2.3 2.2 2.1
1987-88 3.3 2.4 3.6 3.3 2.6 4.0 2.9 1.4 3.4 2.6 2.9 3.3 2.0 1.8 2.4
1992-93 3.9 3.1 5.5 4.1 3.4 5.4 4.8 2.7 5.7 2.9 2.6 4.4 2.6 2.8 3.5
1993-94 4.9 3.2 7.4 3.7 3.5 5.5 4.0 2.2 5.3 3.6 3.0 6.5 3.2 2.5 4.2
1996-97 4.6 4.0 4.3 4.4 4.6 3.5 3.9 3.1 4.3 3.5 3.3 4.2 3.8 2.1 3.1
1998-99 7.3 5.0 7.2 6.8 4.7 7.5 5.7 4.9 6.2 7.4 4.8 6.6 5.4 4.5 3.9
2001-02 4.3 2.3 12.4 5.0 3.3 9.9 4.9 1.3 20.9 2.0 1.3 10.1 2.7 2.0 6.6
2004-05 4.8 2.1 17.2 4.6 2.4 13.8 7.8 1.7 33.2 2.4 1.5 9.5 3.5 2.3 9.3
2005-06 5.9 3.2 31.6 8.2 5.0 34.9 4.1 0.7 36.6 3.6 2.8 15.2 1.2 0.8 5.8
2007-08 5.9 3.4 22.5 6.7 4.6 17.8 5.3 0.8 57.2 5.0 3.7 21.6 1.0 0.3 6.3
Source: Own calculations till 1998-99 using data from HIES/PSLM (various issues)
80
4.3.10 Human Capital Index (HCI)
In addition to physical capital, human capital (HC) is now considered to an
important determinant of economic growth and well-being of the masses. Human capital
is a comprehensive concept that comprises good health and physique and the level of
education, training, acquisition of technical skills and experience. The economists
consider the expenditure made or the funds allocated to education, training and medical
care as investment in human capital (HC).
It is difficult to measure human capital or to compare it across individuals or
different regions. However, the level of health and education are conventionally
considered as significant determinants of HC. The present study follows the same
convention and attempts to build an index for use in the analysis. In this context, we
follow the convention used in the construction of Human Development Index (HDI) vide
the Human Development Report (UNDP-1997).
The main indicators used are the educational attainment and health status. The
educational attainment index captures the effects of literacy rate and combined
enrollment rate, whereas the health index includes the effects of both infant survival rate
and crude birth rate that determine the life expectancy at birth.
(a) Educational Attainment Index (EAI)
Educational attainment index (EAI) is calculated by using zero as a minimum
level and 100 percent as a maximum level of education attainment. Here two third
81
weight is assigned to percentage of literates in labor force (denoted by l ) and one
third weight to combined enrollment rates (denoted by e ).
2 3 1 3
100
l e
EAI
(b) Health Status Index [HSI]
The second part of the human capital is health status, which is measured by the
life expectancy at birth. This indicator has appeared to be significant in many cross
country growth analyses (Bloom and Canning 2000, 2001). Due to the non availability of
relevant data for different regions of Pakistan, we adopted an indirect route. The infant
mortality rate (IMR) and the crude birth rate (CBR) are assumed to be good
determinants of life expectancy at birth. For this purpose, we took data on the three
indicators from 1960 to 2007(panel data set, total 73 observations) for the South Asian
region (Bangladesh, Bhutan, India, Maldives, Nepal, Pakistan, and Sri Lanka) and fitted
the following regression:
77.3416 0.1538( ) 0.15923( )Life Expectancy CBR IMR
By putting the values of crude birth rate and infant mortality rate we obtained the
life expectancy for different regions of Pakistan. To obtain health status index, the
standard procedure given in UNDP, Human Development Report (1997) is followed.
The minimum expected life is 25 years and the maximum is 85 years.
tan 25
85 25
Life Expec cyHSI
82
(c) Human Capital Index
Human capital index (HCI) is obtained by simply taking the average of
educational attainment index (EAI) and health status index (HSI): HCI2
EAI HSI
Using the above formula for the computation of human capital index, the relevant
indicator for different regions of Pakistan is obtained, as shown in Table 4.10. The
relevant information shows that conditions in urban areas are better in this regards than
the rural areas .This trend is also present in the initial period and remains persistent
during the whole period of the study.
The obvious reason is the higher literacy rate in the urban areas, which imparts
awareness in the individuals about health and education. Moreover and as discussed
earlier, there is a visible difference in the income and expenditure of the people residing
in the urban and rural areas. Further the better facilities for health and education depict
better results for urban areas in terms of human capital. The table shows this significant
difference, where the mean human capital index for rural and urban areas of Pakistan is
0.46 and 0.64 respectively.
The overall regions of provinces with reference to human capital index may be
ranked in the sequence: Sindh, Punjab, Khyber Pakhtoonkhwa (KPK), and Balochistan.
This ranking also shows that the urbanization is the major determinant of this variable.
Difference between the last and the initial values in the different regions shows higher
rate of catch up in the rural areas of Pakistan as compared to urban areas. At provincial
level, this difference is not persistent. For instance, Punjab and KPK followed the same
trend as that of Pakistan. However, KPK rural area showed significant progress in
83
human capital where in 1979 value of HCI was 0.33 and in 2007-08 it appeared as 0.58.
These facts are also depicted in the following figures.
84
Table 4.10 Human Capital Index
PAKISTAN PUNJAB SINDH KPK BALOCHISTAN
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 0.42 0.35 0.51 0.41 0.35 0.51 0.45 0.35 0.54 0.39 0.33 0.48 0.40 0.35 0.49
1984-85 0.42 0.35 0.54 0.42 0.36 0.54 0.43 0.32 0.55 0.42 0.38 0.48 0.30 0.26 0.47
1985-86 0.45 0.40 0.58 0.45 0.40 0.58 0.47 0.37 0.60 0.44 0.41 0.55 0.35 0.32 0.48
1986-87 0.48 0.42 0.61 0.48 0.42 0.60 0.48 0.40 0.61 0.46 0.42 0.56 0.38 0.34 0.52
1987-88 0.49 0.42 0.62 0.49 0.43 0.60 0.51 0.38 0.64 0.49 0.45 0.58 0.39 0.35 0.56
1992-93 0.53 0.47 0.68 0.51 0.45 0.65 0.54 0.40 0.69 0.48 0.45 0.63 0.40 0.35 0.60
1993-94 0.54 0.48 0.69 0.53 0.48 0.67 0.56 0.42 0.72 0.53 0.51 0.67 0.44 0.41 0.63
1996-97 0.55 0.49 0.69 0.55 0.50 0.68 0.55 0.44 0.71 0.55 0.53 0.65 0.50 0.48 0.65
1998-99 0.57 0.52 0.68 0.56 0.51 0.69 0.56 0.47 0.72 0.57 0.55 0.68 0.50 0.48 0.67
2001-02 0.54 0.50 0.63 0.57 0.52 0.66 0.57 0.49 0.69 0.54 0.52 0.64 0.48 0.45 0.62
2004-05 0.57 0.51 0.70 0.56 0.51 0.68 0.57 0.47 0.68 0.56 0.55 0.63 0.50 0.46 0.64
2005-06 0.56 0.51 0.68 0.61 0.56 0.72 0.58 0.47 0.71 0.55 0.53 0.62 0.49 0.46 0.59
2007-08 0.58 0.53 0.68 0.59 0.55 0.68 0.58 0.49 0.70 0.59 0.58 0.67 0.54 0.50 0.66
Source: Own Estimation
85
Human Capital Index for overall areas is plotted against years in Figure 4.1,
where on the average, HCI is 0.41 at the initial level and in 2007-08 this value is 0.57.
Figure 4.2 shows that the average HCI for rural areas is 0.34 for the year 1979 and 0.53
for 2007-08. Figure 4.3 reveals that in 1979 average human capital index was 0.51 and
in 2007-08 this value rises to 0.68.
Figure 4.1 Human Capital for Overall Regions
The above figure depicts the situation of overall provinces of Pakistan. There is a
persistent increase from 1979 to 1998-99. Initially, there is no wide gap among regions,
but just after 1979 overall Balochistan took downward jump and remained below the
average throughout the study period. After 1984-85, there is an upward trend that
persists till 1996-97. After that, mixed trend is observed, where some values increased
while others showed downward trend.
86
Figure 4.2 Human Capital for Rural Regions
In the above figure, human capital index for the year 1979 indicates almost the
same standard of health and education in rural section of the country. In general, all
regions observed upward trend overtime, but after 1996-97 the rural Balochistan and
rural Sindh showed stagnant trend, whereas rural Punjab and rural KPK followed more
or less the previous trend. For most of the time, human capital index in rural Balochistan
remained low as compared to other regions, whereas HCI in rural KPK remained
dominant throughout the period.
Figure 4.3 depicts the level of human capital index in urban areas. As expected,
the urban areas are achieving higher level of human capital than the rural areas. Here
urban Sindh dominates throughout (due Karachi metropolitan); which is a clear
indication that standard of health and education in urban Sindh remained well
87
throughout the study period. Like rural areas, the urban regions also show upward trend
till 1998-99. After 1998-99 almost all the regions show slight downward trend especially
in case of urban KPK and urban Balochistan. However both these regions catch up with
urban Punjab and urban Sindh in 2007-08.
Figure 4.3 Human Capital for Urban Regions
4.4 Construction of Basic Needs Gap Index (BNGI)
The basic needs gap index (BNGI) is a measure of poverty level in that it shows
the gap or deprivation level or the distance at which the poor households are standing
before they could attain the objective of basic needs fulfillment. This has been used by
different researchers as indicator and yardstick for policy makers to think of appropriate
88
strategies to attain the objective of basic needs fulfillment (BNF), for instance by Hassan
(1997) and Kipanga (2007). BNGI represents the dependent variable, which is
measured by the following formula:
ptntBNG (4.1)
This measures the difference between the mean expenditure on basic needs nt in the
region and the mean income of the poor in that region, pt .
When equation (4.1) is expressed as a ratio of nt , this gives the index:
ntptntptntBNGI /1][ (4.2)
Generally the income of the poor is less than the mean expenditure on basic needs
( pt nt ) and the index lies between zero and unity. However, in rare cases, it may
happen that income of the poor household exceeds the expenditure on the basic needs
( pt nt ). In such a situation, BNGI will be less than zero. In the present study, it
factually happened. Out of 104 observations for rural and urban areas, we have three
values less than zero. The smaller value of index indicates the better performance; and
the higher value of index shows the worst condition of the region.
To evaluate the mean income of the poor ( pt ), the headcount index as a
measure of poverty ( ) is used for finding the income share ( ) going to a proportion
of population concerned. This can be estimated by using the tables of percentage
distribution of monthly income among the households by quintiles. The following
relationship gives the mean income of the poor.
89
pt Y
. (4.3)
The parameter (α) can be estimated by the following formula,
1 2 1 /L L L n a (4.4)
Where 1L and 2L are the lower and upper limits of the income for population in poverty
respectively. Here n is the value that varies with (people in poverty). If it is above 20
and below 40, the value of n will be 20, and if the value of is more than 40, n will be
40 and so on. Here a assumes the value of 20 if the limits are quintiles; and 10 if the
limits are deciles.
Now to evaluate nt , we need two parameters. The parameter (a ) is the
proportion of consumption expenditure (C ) to household income (Y ), i.e. /a C Y .
The parameter (b ) is the proportion of expenditure on basic needs ( B ) to total
consumption expenditure (C ), i.e. /b B C . Thus we have the equation:
nt tabY (4.5)
Keeping in view the above, the construction of basic needs gap index requires a
lot of information. Some of the data may be readily available, whereas in most cases
one has to evaluate different variables and construct certain parameters. We discuss
the necessary data before we could construct the key variable of our study.
4.4.1 Specification of Basic Needs (B)
Household expenditure is the amount spent by the household on goods and
services for consumption. It also includes own produced goods and those received in
90
kind as remuneration. Household consumption expenditure can be classified as “paid
and consumed” and “unpaid and consumed”, like the wages and salaries received in
kind and consumed, the own produced goods and consumed, and the receipts in the
from of assistance, gifts, dowry, inheritances and other sources.
Published data of HIES includes the following main items under consumption
expenditure:
i) food beverages and tobacco vii) fuel and lighting
ii) apparel, textile and footwear viii) education
iii) transport and communication ix) medical care
iv) cleaning, laundry and personal appearance x) religious functions etc
v) recreation and entertainment xi) litigation expenses
vi) rent xii) miscellaneous expenditure
Out of the above, the following four (4) items are selected as the basic needs and
which constitute about 80% (plus) of expenditure in the poor households:
i) Food, Beverages and Tobacco iii) Rent, Fuel and Lighting
ii) Apparel, Textile and Footwear iv) Health and Education
The average monthly expenditure per household for overall Pakistan and the four
provinces with rural urban bifurcation is shown in Table 4.11.
The average monthly expenditure per household is synchronized with average
monthly income per household. Oscillation in average monthly expenditure reveals that
marginal propensity to consume (MPC) of general public remains more or less the
same. Sharp rise in income in later parts of 2000s can be attributed to seemingly
prudent macroeconomic policies of the military-led regime that could not be sustained
91
by the present government. The year 2007-08 shows downward trend in both income
and expenditure in almost all areas except urban Balochistan. It is apparent from the
Table that in case of Punjab, both income and expenditure increased up to 1986-87, but
after that both the variables oscillated and could not gain momentum.
For the purpose of this study, some information about expenditure on education
and health is needed. The requisite data set on health expenditure and medical care is
not available as a separate entity. The same is therefore extracted from the
miscellaneous expenditures, where expenditure on medical care is given as percent of
miscellaneous expenditure. To obtain the data on healthcare, the ratio of miscellaneous
expenditure to total household expenditure could be multiplied by percent fraction of
miscellaneous expenditure allocated to medical care:
Health as % of total expenditure = [(misc exp/total exp) × % of medicare in misc exp]
As far as the case of education is concerned, the expenditure going to education
is available for the 1986-87 onward, but there is no separate entry in HIES for
expenditure on education before this year. This data is available in column “personal
effects” for the year 1979 and is available in column “entertainment, recreation and
education” for the years 1984-85 and 1985-86. First, we estimated the ratio of both,
“expenditure on education” and “expenditure on recreation” to the gross “expenditure on
personal effect” and using this ratio, we estimated the expenditure on education for the
years 1979 to 1985-86. From 1986-87 onward the data for “expenditure on education” is
available as a separate variable.
Education as % of total expenditure = [(Personal effect/total exp) × % of education in
personal effect or entertainment exp].
92
Table 4.11 Average Monthly Expenditure / Household (Rs) CPI 2005=100
PAKISTAN Punjab SINDH KPK BALOCHISTAN
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 6343 5356 7922 5963 5252 7393 7260 5244 8797 6716 6098 8053 6106 4855 7739
1984-85 7189 6294 9521 6932 6269 8765 8095 6014 10601 7506 7055 10525 5794 5535 8647
1985-86 7375 6388 9930 6981 6236 9157 8609 6423 11343 7423 7192 8850 6267 5819 9063
1986-87 7904 6839 10408 7425 6513 9966 9286 7279 11714 7973 7661 9701 7236 6935 8880
1987-88 7515 6515 10134 7200 6381 9622 8369 6154 11167 7765 7510 9159 6809 6423 9482
1992-93 8105 7181 9865 8102 7400 10119 8874 6756 11777 7447 7288 8423 6160 5835 8842
1993-94 7814 6596 10847 7698 6773 10190 8718 6000 12241 7333 6998 9218 6069 5769 8827
1996-97 7727 6722 10085 7688 6891 9784 8376 5976 10981 7258 7045 8506 6264 5773 8346
1998-99 8559 7257 11609 8015 6789 11065 9559 7260 12428 8641 8068 11872 9923 9709 11491
2001-02 8089 6947 10840 7510 6645 9687 9528 7028 13039 8083 7758 9940 8290 7864 10418
2004-05 9121 7712 12079 8857 7696 11383 10013 7415 13351 8724 8137 11715 8515 7870 11201
2005-06 9901 8282 12960 9849 8256 13258 10310 7459 12825 10181 9752 12358 7587 6890 10189
2007-08 9042 7948 11144 9007 8061 11040 9086 6839 11310 9622 9200 11619 7616 6661 10244
Source: HIES/PSLM (various issues)
93
Table 4.12 Average Monthly Income / Household (Rs) CPI 2005=100
PAKISTAN PUNJAB SINDH KPK BALOCHISTAN
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 6880 5571 8973 6372 5519 8090 7872 5296 9838 7656 6243 10712 6800 5080 9043
1984-85 7714 6686 10392 7422 6637 9496 8493 6167 11292 8478 7695 13713 6401 6113 9560
1985-86 7869 6826 10573 7500 6740 9719 9042 6662 12019 7728 7403 9727 7272 6755 10496
1986-87 8249 7099 10955 7706 6710 10262 9652 7457 12282 8288 7945 10188 8568 8256 10273
1987-88 7891 6724 10949 7522 6592 10275 8873 6216 12228 8009 7678 9819 7678 7198 10993
1992-93 8349 7140 11572 8495 7567 11163 9037 6314 12774 7009 6672 9116 6740 6328 10147
1993-94 7990 6629 11369 8018 6965 10853 8622 5612 12522 7308 6894 9637 6312 5890 10208
1996-97 8612 7803 10512 8918 8364 10373 8890 6799 11143 7390 7151 8787 6875 6266 9455
1998-99 9049 7573 12512 8675 7268 12181 9988 7556 13021 8301 7557 12503 10752 10472 12799
2001-02 8636 7267 11933 8249 7120 11092 9728 7119 13392 8219 7631 11577 9283 8567 12853
2004-05 9685 7929 13371 9488 7941 12854 10413 7467 14196 9395 8516 13879 8849 7980 12470
2005-06 11413 10119 13859 11400 10365 13615 12066 9777 14084 11369 10464 15966 8194 7414 11102
2007-08 10326 9018 12836 10430 9540 12341 10585 7436 13703 10031 9451 12783 8125 6689 12081
Source: HIES/PSLM (various issues)
94
4.4.2 Household Income (Y)
Household income is the sum of monetary income and income “in kind”. The
household income comprises receipts from all sources like wages and salaries,
share/rent received from agricultural land/crop, income from farming and crop
production (like the rents earned on agricultural equipments: tractors, tube wells, animal
carts etc), income from livestock farming, poultry sold or slaughtered for domestic, bee
hives and fishery, the income earned as rents from non-agricultural property and leasing
of equipments, transfer receipts and assistance which includes Zakat, insurance claims,
pensions, gifts and grants etc. and all other sources including sale of assets, domestic
remittances, foreign remittances etc.
Average monthly income per household for different regions as well as for rural
urban segments is shown in the Table 4.12. It shows that average monthly income has
increased for the year 1984-85, 1985-86, 1986-87, 1990-91, 1998-99, 2004-05 and
2005-06; whereas it has declined for the remaining years. This rise was very sharp in
2005-06. Average monthly income per household for rural areas portrays the same
pattern as for the overall Pakistan. However, the urban areas show a consistent trend
which is not observable for the rural areas of Pakistan.
The income gap between the rural and urban areas is visible, and which
remained permanent feature over the whole study period. One of the possible reasons
might be the selective privatization and liberalization of the economy, leading to a
sustained increase in GDP of around 6 % per annum during the 1980‟s decade. The Zia
ul Haq regime and the successive governments emphasized the development of rural
sector and backward areas by distributing funds through the elected representatives.
95
4.4.3 Distribution of Income (Construction of Quintiles)
Income distribution is considered as an important factor of social integration by
the researchers. For the construction of quintiles, the data on monthly income of
different classes is derived from different issues of HIES. The first quintile is computed
by taking the ratio of cumulative income and cumulative population to obtain monthly
income for bottom 20%. The second, third, fourth, and the highest quintiles are
constructed. For some values of α in the formula for finding BNGI, we need the first
decile or income share going to the bottom 10% population. Quintiles after the year
2000 are given in the HIES published data.
Table 4.13 shows income distribution by quintiles for overall, rural and urban
Pakistan. The overall picture shows a dismal position where the distribution is getting
more unequal overtime. The share of income for the first quintile goes down by 33%
(12.2 to 7.9) over the period 1979-90 to 2007-08, where this decrease is 26% and 71%
respectively for the rural and urban areas. However, this decrease is very large and
consistent in case of urban areas, whereas in the rural areas of Sindh and Balochistan
an increase in the share of income is observed .This change is more vivid after 2005-
06. The highest quintile (top 20% share of income) shows an increase overtime, except
for rural Sind, overall and rural KPK, and rural Balochistan.
So, we can conclude that the poor is getting poorer and the rich getting richer
overtime. There is an increase in share of income held by the highest quintile in all the
cases. This increase is 22% for the overall Pakistan; and 18% and 51% respectively for
rural and urban areas.
96
Table 4.13 Percentage Distribution of Monthly Income Among Households
by Quintiles PAKISTAN
OVERALL
RURAL URBAN
Years 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th
1979 12.2 14.1 15.6 19.9 38.2 14.1 16.4 17.6 21.7 30.2 10.6 12.3 14.0 18.2 45.0
1984-85 9.7 13.8 15.8 17.3 43.3 12.4 16.7 18.6 19.1 33.3 10.8 13.4 14.3 15.7 45.7
1985-86 12.0 14.8 16.8 18.4 38.1 13.2 16.3 18.1 19.6 32.8 10.4 12.8 14.9 16.5 45.4
1986-87 11.8 15.0 16.5 17.8 39.0 13.2 16.7 18.0 19.2 32.9 10.1 12.5 14.7 15.8 46.8
1987-88 12.0 14.6 16.3 17.7 39.4 13.4 16.3 18.3 19.4 32.6 12.0 13.7 15.2 16.0 43.0
1990-91 9.2 12.2 16.3 15.7 46.6 4.3 11.3 16.5 23.2 48.9 10.2 14.1 14.5 15.3 45.9
1992-93 11.7 13.1 14.6 15.4 45.2 12.8 14.4 15.9 16.6 40.2 9.7 11.0 12.6 13.8 53.0
1993-94 9.4 11.9 16.1 16.9 45.7 11.5 15.3 16.2 20.6 36.4 7.5 11.6 11.8 13.7 55.4
1996-97 10.6 11.8 12.8 16.2 48.6 11.3 12.5 13.5 16.9 45.7 10.9 13.0 14.1 15.2 46.8
1998-99 7.9 9.1 11.7 13.2 58.0 9.9 11.1 14.1 15.3 49.5 8.2 9.4 11.1 12.3 59.0
2001-02 9.7 12.7 15.7 20.3 41.7 13.0 16.4 18.8 22.3 29.6 4.8 7.2 11.2 17.3 59.5
2004-05 9.3 11.9 14.5 19.7 44.6 13.8 16.9 18.2 22.0 29.2 3.7 5.8 9.9 16.9 63.8
2005-06 7.6 11.4 15.4 20.6 45.0 10.4 14.3 18.6 23.5 33.2 2.2 5.5 9.1 14.8 68.5
2007-08 7.9 11.2 14.5 19.7 46.8 10.5 13.9 17.5 22.5 35.6 3.0 6.1 8.7 14.3 67.8
Source: Own calculations till 1998-99 using HIES/PSLM data (various issues) .
97
Table 4.14 Percentage Distribution of Monthly Income Among Households
by Quintiles PUNJAB
OVERAL
RURAL URBAN
Years 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th
1979 12.1 14.6 16.1 20.8 36.3 13.5 16.1 17.5 22.0 31.0 10.6 12.8 14.6 19.1 42.9
1984-85 10.3 14.3 16.6 18.1 40.7 12.4 16.3 18.5 19.2 33.6 10.3 13.6 15.2 17.2 43.7
1985-86 12.0 14.9 17.1 18.5 37.5 12.9 16.0 18.2 19.5 33.4 10.7 12.6 15.0 16.8 44.9
1986-87 12.1 15.1 16.9 18.1 37.9 13.4 16.8 18.2 19.5 32.0 10.1 12.3 14.9 15.8 46.9
1987-88 11.9 14.6 16.3 17.9 39.3 13.0 16.0 18.1 19.5 33.4 11.4 12.2 14.3 16.0 46.1
1992-93 11.4 12.8 14.5 15.2 46.1 12.3 13.9 15.6 16.0 42.2 9.7 11.1 12.8 14.0 52.4
1993-94 11.3 13.7 15.4 17.2 42.3 10.9 13.1 17.5 20.9 37.6 9.5 10.2 11.1 16.8 52.3
1996-97 10.8 11.8 13.8 15.9 46.9 10.7 11.6 12.8 15.9 49.1 10.8 20.8 15.2 16.0 37.2
1998-99 8.4 9.1 12.3 13.8 56.4 10.2 11.7 12.6 17.2 48.2 8.0 9.2 10.9 11.7 60.1
2001-02 8.8 11.9 15.1 20.7 43.5 10.6 14.7 17.2 23.1 34.5 5.9 7.3 11.8 16.9 58.2
2004-05 9.9 11.4 13.5 19.5 45.7 14.0 15.3 15.4 21.6 33.8 4.5 6.3 11.0 16.6 61.7
2005-06 5.9 9.4 15.3 21.0 48.4 7.7 11.3 18.3 24.4 38.3 2.0 5.1 8.8 13.6 70.5
2007-08 7.5 10.0 13.7 18.3 50.5 9.1 11.7 15.9 21.9 41.5 4.0 6.4 8.7 10.0 71.0
Source: Own calculations till 1998-99 using HIES/PSLM data (various issues)
98
Table 4.15 Percentage Distribution of Monthly Income Among Households
By Quintiles SINDH
OVERALL
RURAL URBAN
Years 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th
1979 12.7 13.2 14.8 18.4 40.9 17.2 17.9 19.0 21.5 24.5 12.0 11.8 13.6 16.9 45.7
1984-85 11.0 13.8 15.3 16.7 43.2 15.3 17.3 19.0 19.9 28.5 13.8 14.9 13.7 13.7 43.8
1985-86 11.6 14.4 16.0 17.7 40.4 14.5 17.3 18.7 20.1 29.4 11.1 13.7 14.0 16.0 45.2
1986-87 12.3 14.2 15.3 17.0 41.2 14.7 17.0 17.8 19.3 31.1 11.7 12.8 13.9 15.6 46.0
1987-88 13.2 14.7 16.1 17.2 38.7 17.3 18.7 19.5 20.1 24.3 12.6 13.9 15.0 16.1 42.4
1992-93 10.4 12.0 13.2 14.7 49.8 13.5 15.3 16.3 17.9 37.0 9.5 10.8 12.1 13.3 54.3
1993-94 11.7 11.8 13.8 16.0 46.7 14.0 18.1 17.7 18.8 31.4 9.9 11.5 12.2 14.1 52.3
1996-97 11.4 12.3 15.3 16.3 44.6 12.8 14.4 14.9 18.4 39.6 10.9 13.2 13.8 15.2 46.8
1998-99 9.4 11.1 12.7 13.4 53.4 10.1 11.3 12.9 15.9 49.9 9.4 9.1 11.6 11.8 58.1
2001-02 9.5 11.1 14.1 18.8 46.6 18.1 17.4 19.6 20.8 24.1 3.0 6.4 9.9 17.2 63.5
2004-05 6.5 9.9 13.7 19.7 50.3 13.0 18.0 22.9 24.1 22.1 2.1 4.4 7.5 16.7 69.4
2005-06 10.3 13.8 16.6 16.8 42.5 18.9 23.3 24.7 20.0 13.1 1.9 4.6 8.8 13.7 71.0
2007-08 8.6 12.2 13.6 19.7 46.0 19.2 23.2 21.7 20.7 15.2 1.2 4.5 7.9 19.1 67.4
Source: Own calculations till 1998-99 using HIES/PSLM data (various issues)
99
Table 4.16 Percentage Distribution of Monthly Income Among Households
By Quintiles KPK
OVERALL
RURAL URBAN
Years 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th
1979 11.7 13.2 14.4 18.5 42.3 13.7 15.6 16.7 20.9 33.2 8.7 10.1 25.0 13.5 42.7
1984-85 8.1 11.5 14.2 14.8 51.4 11.2 15.4 17.7 17.4 38.2 8.9 10.1 12.7 13.4 54.9
1985-86 13.0 15.6 16.9 18.5 36.1 13.4 16.2 17.3 19.0 34.0 10.8 12.9 15.1 16.0 45.2
1986-87 11.5 15.6 17.0 17.8 38.1 11.7 16.2 17.3 18.3 36.5 11.2 12.8 15.8 15.5 44.7
1987-88 13.5 16.1 17.0 17.7 35.7 12.4 15.2 17.8 18.2 36.4 12.7 14.3 15.5 15.9 41.5
1992-93 13.4 15.1 16.2 16.9 38.5 14.1 15.8 16.8 17.3 36.1 10.7 12.2 14.1 15.4 47.5
1993-94 10.7 15.8 16.7 19.0 37.9 12.0 15.3 16.1 20.9 35.7 7.9 12.3 13.1 15.5 51.2
1996-97 12.1 13.2 14.2 18.5 41.9 12.4 13.6 14.4 19.2 40.4 11.4 11.8 13.3 15.5 48.0
1998-99 8.1 9.5 10.3 12.7 59.4 10.3 12.3 12.7 15.0 49.6 8.8 10.4 9.9 12.4 58.5
2001-02 14.0 18.4 19.4 20.7 27.5 16.3 20.5 21.5 20.9 20.8 5.2 10.7 11.4 20.1 52.7
2004-05 12.9 18.6 19.4 18.3 30.8 15.1 21.6 21.7 18.5 23.1 5.8 9.4 12.1 17.9 54.9
2005-06 10.2 15.8 13.4 23.7 36.9 11.9 17.5 14.1 23.2 33.4 3.4 9.3 10.5 26.0 50.9
2007-08 6.9 14.0 18.8 26.3 34.1 7.9 15.8 21.1 26.1 29.2 2.6 6.3 8.9 27.1 55.2
Source: Own calculations till 1998-99 using HIES/PSLM data (various issues)
100
Table 4.17 Percentage Distribution of Monthly Income Among Households
By Quintiles BALOCHISTAN
OVERALL
RURAL URBAN
Years 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th 1st 2nd 3rd 4th 5th
1979 13.3 14.6 16.5 21.0 34.6 16.0 17.5 19.3 21.3 25.8 12.0 12.5 15.1 21.3 39.1
1984-85 8.7 16.1 15.9 19.2 40.1 11.6 19.4 18.6 19.1 31.3 13.0 15.9 16.2 19.6 35.3
1985-86 12.8 14.9 16.9 20.2 35.2 13.3 15.5 17.3 20.9 32.9 12.2 13.2 15.3 17.7 41.6
1986-87 15.5 14.2 16.9 18.1 35.3 15.6 14.1 17.1 18.1 35.1 16.9 15.0 16.0 17.4 34.7
1987-88 15.3 16.7 18.5 19.0 30.4 15.8 17.3 19.1 19.2 28.5 14.5 15.5 17.2 17.4 35.4
1992-93 13.6 15.3 17.1 18.4 35.6 13.0 14.7 16.8 19.8 35.8 12.2 13.5 13.6 17.4 43.3
1993-94 11.4 14.1 16.4 21.1 37.0 13.0 15.6 17.7 21.4 32.3 10.3 12.3 14.2 20.5 42.7
1996-97 10.8 13.6 15.5 18.9 41.1 14.8 16.1 17.4 20.0 31.7 13.0 14.7 15.7 16.8 39.8
1998-99 10.2 10.8 12.0 12.0 54.9 11.8 11.2 12.0 11.8 53.1 11.6 13.9 13.0 15.9 45.6
2001-02 10.2 16.8 22.9 22.6 27.6 11.2 19.0 24.9 23.1 21.9 7.1 9.3 16.1 21.1 46.5
2004-05 8.8 13.8 18.6 28.2 30.5 10.2 16.3 20.8 29.4 23.4 5.3 7.3 12.8 25.0 49.6
2005-06 17.2 19.3 20.3 23.1 20.2 20.2 20.5 21.3 22.7 15.3 6.4 15.3 16.5 24.5 37.3
2007-08 17.4 22.8 20.9 21.8 17.1 24.3 23.2 21.7 23.5 7.3 5.4 22.3 19.6 18.7 34.0
Source: Own calculations till 1998-99 using HIES/PSLM data (various issues)
101
Table 4.14 depicts the situation of Punjab, which is more or less similar to that of
overall Pakistan. In case of Sindh rural as shown in Table 4.15, the results are unusual
where the data shows 12% increase in the percentage share of income of the poor over
time and 38% decrease in the highest quintile. The middle three quintiles share more
than 65% of income and the highest quintile shows very unusual low values of 38% and
15% for the years 2005-06 and 2007-08 respectively.
In Sindh urban case, we have the opposite trend in the first quintile with a huge
decrease in the income share i.e. 12% in1979 to 1% in 2007-08. On the other hand,
there is an increase of income share by about 8% in the highest quintile. Although all
the urban areas show a wide gap between income shares of the poor and the rich; but
this gap is very large in case of Sindh urban. The reason might be that urban Sindh
includes Karachi, the largest trading and industrial city of Pakistan, where the middle
class is gaining weight.
The percentage distribution of monthly income by quintiles for KPK is given in
Table 4.16. The income share of both the first and the fifth quintile have decreased for
overall KPK, while the share of the middle quintiles has increased overtime. KPK urban
shows a huge decline in the share of income held by the first quintile i.e. 70% decrease,
whereas an increase in the share held by the highest quintile is observed by 29%.
Balochistan portrays a different picture when it is compared with other regions of
Pakistan. Table 4.17 depicts the picture of income distribution. Balochistan overall
shows an increase in the income share of poor by 30% but the highest quintile shows a
decline by 50%. Year 2007-08 shows a value of 17.13 for the highest quintile, which is
102
quite unusual when compared to other parts of the country. Balochistan rural presents
more unusual results and the situation seems just opposite when we compare the
values of the initial and final years. Here, the 1st quintile shows a decrease from 16.0 to
10.2 up to the year 2004-05 and then a sudden increase to 24.3 by the year 2007-08.
Likewise, the highest quintile shows somewhat constant but fluctuating trend up to
2004-05 but a sudden drop from 23.4 to 7.28 by the year 2007-08. This may be due to
problems in data collection and compilation on account of security issue and
deteriorating law and order situation. Another probable reason behind this unusual
change might be the policies adopted by the Musharaf regime; where funds were
generously provided by the federal government for the least developed areas of
Balochistan.
4.4.4 The Poverty Status of Households (The Headcount Ratio)
One component that is used during the construction of BNGI is Head Count Ratio
for the different regions of Pakistan. Data about head count ratio, especially at provincial
level with rural urban bifurcation, is not available from a single source. The data on
poverty indices is available from early 1990‟s onward. Prior to this period, there were
certain individual studies that attempted to estimate the poverty indices; which is surely
a great contribution in this area of research. These include Amjad.R and Kemal. A.R
(1997), Cheema. I. A. (2005), Irfan. M. (2007), Jamal.H. (2006), Qureshi.S.K and
ARIF.G.M, (2001),Ellahi Mahboob, Khan S.R. Rafi (1999), Shirazi.N.S (1993), Zaidi S.A
(2000).
As mentioned earlier, the HIES data for 1990-91 is only for all Pakistan level,
therefore we have one extra entry. For the year 1985-86 and 1986-87, we could not find
103
data on the headcount index and it is interpolated by using weighted average formula2.
Data for the last two entries is obtained by using the 5-year moving average.
Table 4.18 shows head count ratio (HCR) for Pakistan and the four provinces
with rural-urban bifurcation. If this measure of poverty is reliable, then the poverty
remains more or less the same over the long study period for overall Pakistan. However
there is a visible difference in urban and rural area. There is a slight increase in HCR in
rural areas but a significant reduction in poverty in urban areas, from 25.9 in 1979 to
19.7 in 2007-08.
Punjab (overall, rural and urban) shows approximately the same situation of
poverty for the year 1979. Total Punjab and rural area shows slight decrease in HCR
figures, whereas a significant reduction in poverty can be seen for urban Punjab. In
case of Sindh, a significant reduction in HCR can be seen for all the three cases (total,
rural and urban). Here the pattern of change is similar to Punjab. The situation in
Balochistan is more or less similar to that of Punjab and Sindh. However, the situation
of poverty in KPK worsened over time both in overall and rural areas. However, the
urban areas of KPK recorded a slight reduction in HCR like other urban regions of
Pakistan.
Generally speaking, the urban areas of Pakistan have enjoyed some reduction in
poverty (HCR), whereas the situation in rural areas either remained same or got
worsened in some cases. The poverty figures measured as the HCR remained low in all
regions of Pakistan for the year 1987-88.
2 0.7 weight to the previous value and 0.3 to next to previous value) in the same way for the year 1986-
87, 0.7 weight to the proceeding value and 0.3 to next to the proceeding value is given.
104
Table 4.18 POVERTY (Head Count Ratio)
PAKISTAN PUNJAB SINDH NWFP BALOCHISTAN
YEARS Total Rural Urban YEARS Total Rural Urban Total Rural Urban Total Rural Urban Total Rural Urban
1979 30.7 32.5 25.9 1979 35.1 35.3 35.1 38.2 40.3 37.5 34.9 35.1 34.5 33.70 36.30 32.00
1984-85 18.3 21.1 11.1 1984-85 19.0 21.3 12.8 15.3 22.2 7.0 9.6 9.0 7.5 27.5 28.5 17.0
1985-86 22 24.5 15.6 1985-86 23.83 25.5 19.49 22.17 27.63 16.15 17.19 16.83 15.6 29.36 30.84 21.5
1986-87 21.8 24.8 14.5 1986-87 21.21 23.43 14.70 13.64 18.79 7.17 20.94 21.67 15.99 14.54 14.86 12.21
1987-88 16.6 19.6 8.7 1987-88 19.9 22.6 11.9 9.5 14.6 3.1 15.5 16.0 12.4 9.3 10.0 4.40
1990-91 34 36.9 28 1992-93 24.25 25.37 21.24 23.29 28.56 16.65 33.62 34.91 24.37 26.77 26.21 30.44
1992-93 25.5 27.6 20.0 1993-94 28.55 32.95 17.01 21.5 30.24 11.33 36.37 38.22 25.31 34.36 36.75 15.62
1993-94 28.2 33.5 15.4 1996-97 24.66 27.84 16.61 15.39 19.22 11.77 40.23 42.36 26.92 37.69 41.61 22.98
1996-97 25.8 30.2 15.8 1998-99 31.62 34.62 24.24 26.01 34 15.57 41.28 43.72 27.13 21.55 21.34 22.94
1998-99 31.1 35.1 21.4 2001-02 32.24 35.86 23.33 35.32 45.07 20.06 41.47 43.61 29.05 35.49 37.45 26.18
2001-02 34.5 39.3 22.7 2004-05 24.3 28 16.3 18.3 23.7 11 32.1 34.1 21.9 26.7 28.8 18.5
2004-05 23.9 28.1 14.9 2005-06 28.21 31.58 20.12 23.75 30.49 14.6 38.77 40.94 26.25 30.35 32.3 22.65
2005-06 28.8 33.2 18.7 2007-08 29.38 32.83 21.29 26.54 34.25 15.54 38.28 40.47 26.026 27.91 29.19 22.54
2007-08 29.8 34.2 19.7
Source: HIES and Miscellaneous sources
105
Table 4.19 Construction of the BNGI (Overall Punjab) Y
EA
RS
AV
G.M
ON
TH
LY
IN
CO
ME
/ H
.H
(Y)
AV
G: M
onth
ly C
ons: E
xp: / H
H (
C )
Cons. E
xp.
as %
of H
H I
ncom
e
a=
C/Y
Fo
od,
bevera
geg &
tobacco
Appare
l, t
extile
and footw
ear
Rent
+F
uel &
Lig
htin
g
Health +
Ed
ucatio
n
Exp:
on B
asic
Needs a
s %
of C
ons:
b
=(B
/C)*
100
People
in
Povert
y a
s %
of
Tota
l
Popula
tio
n λ
(H
C)
Lim
its o
f In
com
e (
%)
for
(%
) of P
op.
in P
ov. (B
ased o
n R
ela
vant Q
uin
tile
s
/ D
ecile
Valu
es)
Estim
ate
d %
of In
com
e
with
Corr
espondin
g %
of P
eople
in
Povert
y α
Ént =
a*b
*Yt
Y¯
pt =
(α/λ
) Y
t
BN
GI
=1
- (Y
¯pt/É
nt)
Y C a=C/Y b Λ L1 L2 α Ént Ýpt BNGI
1979 6372 5963 0.94 50.6 10.0 15.5 2.0 0.78 35.1 12.1 14.6 14.0 4656 2544 0.45
1984-85 7422 6932 0.93 47.9 7.7 16.3 3.9 0.76 19.0 4.4 10.3 7.1 5257 2756 0.48
1985-86 7500 6981 0.93 47.4 7.8 16.9 3.8 0.76 23.8 12.0 14.9 14.0 5293 4409 0.17
1986-87 7706 7425 0.96 45.5 7.6 17.4 3.6 0.74 21.2 12.1 2.0 6.4 5499 2335 0.58
1987-88 7522 7200 0.96 45.1 8.0 17.7 3.7 0.75 19.9 5.8 11.9 8.9 5366 3346 0.38
1992-93 8495 8102 0.95 48.3 8.7 20.1 4.1 0.81 24.3 11.4 12.8 11.7 6581 4091 0.38
1993-94 8018 7698 0.96 48.6 8.6 20.2 4.7 0.82 28.6 11.3 13.7 12.4 6325 3471 0.45
1996-97 8918 7688 0.86 48.5 8.5 19.7 4.2 0.81 24.7 10.8 11.8 11.0 6224 3989 0.36
1998-99 8675 8015 0.92 47.8 8.4 21.8 8.5 0.86 31.6 8.4 9.1 8.8 6928 2416 0.65
2001-02 8249 7510 0.91 47.9 7.2 21.5 7.9 0.85 32.2 8.8 11.9 10.7 6347 2734 0.57
2004-05 9488 8857 0.93 48.0 9.2 20.9 7.1 0.85 24.3 9.9 11.4 10.2 7549 4001 0.47
2005-06 11400 9849 0.86 42.9 6.2 23.1 7.8 0.80 28.2 9.4 15.3 11.8 7874 4773 0.39
2007-08 10430 9007 0.86 43.5 5.9 22.6 7.7 0.80 29.4 10.0 13.7 11.7 7174 4167 0.42
Source: Own Estimation
106
4.4.5 The Construction of BNGI
Table 4.19 shows the procedure for and construction of the dependent variable
BNGI, taking the data of overall Punjab. The first column contains years whereas the 2nd
and 3rd columns show the average monthly income and average monthly consumption
per household denoted by Y and C respectively. To find the mean expenditure on basic
needs nt , we need a and b . Column 4 shows consumption expenditure as percentage
of household income which is represented by „α‟ while column 5 through 8 contains
expenditure on basic needs that is (1) food, beverages and tobacco, (2) apparel, textile
and footwear, (3) rent, fuel and lighting, (4) health and education as percent of total
consumption. Sum of the above entities is given in column 9 and is denoted by „b‟.
The people in poverty (as % of total population i.e. head count ratio) is
denoted by „λ‟ and is given in column 10. Lower L1 and upper Limits L2 of Income
against population in poverty, both expressed in percentage is mentioned in column
11. In column 13, the parameter „α‟ is estimated as percent of income with
corresponding percent of people in poverty using Equation 4.4. nt and pt are
calculated by using formulas contained and shown in column 13 and 14. The last
column contains values of BNGI, as dependent variable.
In the same way BNGI for overall, rural and urban areas of Pakistan and each
province is constructed and the values for all these regions are given in Table 4.20.
The analysis follows in the next section.
107
Table 4.20 BNGI for all Regions of Pakistan
PAKISTAN PUNJAB SINDH KPK
BALOCHISTAN
YEARS Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban Overall Rural Urban
1979 0.41 0.38 0.38 0.45 0.41 0.51 0.52 0.45 0.54 0.48 0.47 0.52 0.42 0.39 0.45
1984-85 0.33 0.12 0.31 0.48 0.16 0.39 0.40 0.08 0.51 0.50 0.28 0.64 0.42 0.27 0.28
1985-86 0.23 0.21 0.41 0.17 0.23 0.44 0.27 0.26 0.14 0.17 0.32 0.65 0.33 0.33 0.18
1986-87 0.23 0.26 0.40 0.58 0.16 0.42 0.29 0.54 0.53 0.28 0.28 0.61 0.01 -0.005 -0.06
1987-88 0.35 0.32 0.15 0.38 0.17 0.31 0.56 0.47 0.56 0.32 0.37 0.23 0.46 0.45 0.49
1992-93 0.40 0.42 0.48 0.38 0.37 0.37 0.42 0.43 0.50 0.53 0.54 0.39 0.34 0.32 0.42
1993-94 0.54 0.50 0.41 0.45 0.53 0.49 0.33 0.40 0.27 0.52 0.55 0.55 0.53 0.51 0.34
1996-97 0.42 0.44 0.44 0.36 0.41 0.47 0.42 0.37 0.37 0.25 0.28 0.47 0.54 0.04 0.21
1998-99 0.66 0.63 0.50 0.65 0.60 0.56 0.53 0.61 0.50 0.53 0.45 0.58 0.40 0.31 0.33
2001-02 0.56 0.49 0.70 0.57 0.51 0.64 0.62 0.04 0.81 0.07 0.03 0.63 0.42 0.37 0.56
2004-05 0.47 0.34 0.70 0.47 0.35 0.76 0.68 0.29 0.86 0.35 0.29 0.60 0.50 0.46 0.72
2005-06 0.54 0.22 0.62 0.39 0.52 0.77 0.31 -0.13 0.88 0.45 0.44 0.87 0.20 0.17 0.55
2007-08 0.54 0.32 0.55 0.42 0.51 0.65 0.46 0.12 0.92 0.55 0.50 0.91 0.10 0.02 0.55
Source: Own Estimation
108
4.4.6 Analysis of Basic Needs Gap Index
The basic needs gap index is given in Table 4.20 for different regions of
Pakistan. The behaviour of BNGI is also depicted graphically. The index for rural and
urban Pakistan is shown in Figure 4.4. The indices emanate from 1979 with a value of
0.38. The gap between rural and urban Pakistan widened immediately in succeeding
years. The index for rural Pakistan fell drastically while that for urban areas surged
modestly. A fair explanation for decline in the rural index is that local elections were held
in 1979. More emphasis was laid on rural development and enormous funds were
distributed through local representatives. This led to an increase in economic activity,
employment opportunities and increase in production.
Zia‟s regime specifically targeted backward and rural areas of Pakistan and
special measures were taken to uplift the less developed regions. According to Tahir et
al (1997), involvement of former USSR in Afghanistan from 1979 to 1987-88 remained
the major reason of foreign capital flows during this period. The economy grew with a
sustained rate of around 6% during the 1980s decade, which caused a reduction in
income inequality. This study further argues that export of manpower to Middle East in
mid 1970s resulted into large remittances, which reached to peak figures in 1982-83,
and contributed to reduction in rural and urban inequality. In 1988 a structural
adjustment program was initiated under the auspices of IMF. Some reforms were
introduced, which included increase in tax rates and withdrawal of subsidy on certain
items, specifically on the agricultural inputs and production. These reforms adversely
affected the rural segments of Pakistan. BNGI fell to its lowest ebb (0.15) in 1987-88 but
109
there is an increasing trend after 1988. It is conservable fact that latter governments
carried out different sort of programs according to their policy priorities.
Some governments wanted to attack poverty through social action programs
while others tried to alleviate it through distribution of funds among the public
representatives. However, no government managed to make dent in poverty through
long run viable and sustained development policies. Five breakeven points are
traceable if one observes the figure. The gap between rural and urban BNGI beyond
2001-02 is continuous, although not uniform, and it lasts till the end of the study period.
BNGI touches 0.7 points during 2005-06 and the gap tends to converge after 2005-06.
Figure 4.4 BNGI. Pakistan Rural and Urban
Same relationship of BNGI between rural Punjab and urban Punjab is traceable
in Figure 4.5. However BNGI line for urban Punjab emanates from 0.5 point and for the
rural Punjab it starts from 0.4 points. Both rural and urban Indices showed downward
110
trend up to 1990. Rural index travels beneath urban index at large up to 1992-93 and it
lies above the urban index between 1992-93 and 1993-94. Rural index falls below urban
index perceptually after 2001-02. This trend resembles BNGI for overall Pakistan urban
and rural areas. The spell ranging from 1987-88 to 1998-99 exhibits quite interesting
results regarding BNGI. The country embarked on the path of democracy during this
spell. Ironically, democratic governments changed frequently without completing their
tenures. It was a severe blow for the country‟s economy because every new political
government rolled back the ongoing projects and changed the direction of their policies
spoiling the process of economic growth. Due to these oscillatory strategies, BNGI for
rural and urban Punjab coincides with each other. The same relationship can be
observed for rural and urban BNGI. During this spell, BNGI kept increasing moderately.
Figure 4.5 BNGI. Punjab Rural and Urban
111
Figure 4.6 BNGI. Sindh Rural and Urban
BNGI for urban and rural Sindh is depicted in Figure 4.6. A parallel fall in urban
and rural BNGI is obvious up to 1984-85. BNGI for both rural and urban Sindh twisted
around 0.4 points. These indices began diverging after 1988-89. Urban BNGI rose
persistently. This index tends to 1.0 point whereas rural BNGI plumbed down and even
it became negative in 2006-07.
In Figure 4.7 BNGI for rural and urban KPK has been portrayed. BNGI for rural
KPK stemmed from 0.5 and it rolled up touching the value of 0.68, afterwards, it turned
down, reaching 0.2, then further showing oscillatory trend fall to zero in 2001-02, but it
started climbing again parallel to urban KPK. BNGI for urban KPK started from 0.5 and
contrary to its rural counterpart it went above rural KPK and remained high till 1985-86.
Then it remained below for the period of 1986-87 to 1993-94. After this, there is
112
divergence till 2001-02. Contrary to other provinces, there is a parallel increase in BNGI
for rural and urban KPK from 2001-02 to 2007-08.
Figure 4.7 BNGI. KPK Rural and Urban
BNGI for rural and urban Balochistan has been displayed in Figure 4.8, where
both rural and urban BNGI chased each other. BNGI constructed for rural and urban
Balochistan has zigzag trend. Both fall up to 1986-87, increase and then again follow
each other. In 1986-87 rural BNGI went almost zero whereas urban BNGI lagged
behind to 0.2. Both indices increase up to 2004-05 and then start decreasing.
Summing this discussion up, the research concludes that the urban BNGI on average
lies above rural BNGI for the study period.
113
Figure 4.8 BNGI. Balochistan Rural and Urban
Very few studies have been done for inequalities at province level and
especially with rural urban bifurcation. Our results are more or less same as that of
earlier studies for the overall areas. However comparison of rural and urban areas
gives us clearer picture.
As mentioned by Idrees (2006) about earlier studies and their results.
“The official estimates do not provide any estimates regarding incidence of
income inequality within provinces of Pakistan. However, Kruijk (1986), Ahmad
and Ludlow (1989), Ahmad (2000) and Talat (2003) are few major efforts in this
regard each of these studies was conducted for shorter period of time. In general
these studies show that in late I970's and early 1980's the income inequality has
been maximum in the province of KPK. In 1990 's the maximum income
114
inequalities prevailed in Sindh followed by Punjab and then NWFP. Baluchistan
in general had the least degree of income inequalities during all surveyed years. ”
115
CHAPTER 5
MODEL SPECIFICATION AND METHODOLOGY
Our preliminary/intuitive model includes ten (10) explanatory variables against
the dependent variable, which is the basic needs gap index (BNGI). The model can be
written in the general format as under:
1
K
it ki kit it
k
Y X
The dependent variable (BNGI) is denoted by „Y‟, where the subscript „i‟ stands for the
region (rural/urban in the province concerned) and „t‟ stands for the time period. The
explanatory variables are generally denoted by „X‟, where the disturbance term is given
by „v‟, with all the standard classical assumptions. Detailed discussion and justification
of the explanatory variables is given in Chapter-4. The model may be written in the
simple linear form as follows (Eq. No 5.1):
0 1 2 3 4 5 6 7 8 9 10Re 20 2it i i it i it i it i it i it i it i it i it i it i it itBNGI YPC SPC m HS HE DR Un B T B HCI V
However, it is possible that some of these variables may be irrelevant or insignificant.
Whether the suggested model is just fitted and whether all the explanatory variables are
theoretically consistent and satisfy the general criterion of “weakly exogenous
regressors”? All these questions need to be answered carefully and the model
specification bias must be avoided. We discuss different possibilities to answer these
questions in the next section.
116
5.1 Model Selection
The procedure for selecting explanatory variables needs proper attention; on one
hand we want our model to be parsimonious and simple, and on the other hand, it
should incorporate all the essential variables since exclusion of any relevant variable
might cause missing variable bias. We adopt general to simple methodology for
selection of appropriate and theoretically relevant variables in the model. Alternatively,
the variables which do not contribute significantly to dependent variable should be
excluded. In order to get an appropriate model, there are number of procedures and all
of which have their merits and demerits. There is no clarity in literature as to which of
the proposed methods is superior. The present study uses three methods to select
appropriate regressors for the model. These methods are;
Descriptive statistics (Correlation),
Static panel, and
Impulse Saturation
The variables that are insignificant by all of these methods are not included in the
model. Alternatively, if a variable is supported by at least one of the three methods, we
will include it in the final analysis. This is because if we drop a relevant variable, it will
result into biased specification. On the other hand, if we include an extra and less
significant variable, the results of model will remain unbiased and consistent. The
details of these methods and the results are summarized in the forthcoming sections.
117
5.1.1 Statistical Analysis
For this study, ten determinants of the basic needs gap have been proposed;
namely the per capita income(Ypc), per capita savings (Spc), remittances both domestic
and foreign (Rem), household size (HS), higher education (HE), dependency ratio (DR),
unemployment rate (Un), share of income going to bottom 20% of population (B20),
ratio of income of top 20% to bottom 20 % (T2B), and human capital represented by the
health and education index (HCI)
To get relevant regressors for the final model, both economic theory and the
statistical significance need to be considered. The relevance and importance of the
proposed explanatory variables has been discussed in the previous chapter on data and
variables. So far as the statistical significance is concerned, we have to study the
correlation among the variables under reference. The common wisdom leads us to
determine as to which variables should be retained and which ones to be excluded.
Obviously, the explanatory variables having high correlation with the dependent variable
should be considered and those which have high correlation among themselves should
be excluded. The results of this investigation are shown below:
(i) Aggregate Rural and Urban Areas
The results are shown in Table 5.1. Looking at the Table, the variables like per
capita income, human capital, ratio of income of top 20% to bottom 20%, share of
income held by bottom 20%, higher education and unemployment show relatively high
value of absolute correlation with BNGI; if the standard is absolute value of r > [0.4].
118
Of these six variables, the per capita income (YPc) is the most important one
since the remaining five variables are also correlated with it. Because of its importance,
it will be inappropriate to exclude this variable. Next we see that B20 is linearly
correlated with HCI and T2B; HE is linearly correlated with HCI, T2B and B20; and Un is
linearly correlated with HCI and B20. If the two variables, namely T2B and HE, are
excluded, the final model will be good fit if Ypc, B20, HCI and Un are incorporated. The
specification will assume the following form:
0 1 2 3 420it i i it i it i it i it itBNGI Y HCI B Un v (5.2)
Table 5.1 Correlations, Means and Standard Deviations
Aggregate Rural and Urban Areas
Variables BNGI Ypc Spc Rem HCI HS T2B B20 HE Un DR
BNGI 1.00
Ypc 0.43 1.00
Spc 0.24 0.61 1.00
Rem 0.22 -0.06 -0.15 1.00
HCI 0.43 0.77 0.31 0.25 1.00
HS 0.07 0.03 0.11 0.39 0.38 1.00
T2B 0.64 0.64 0.35 0.04 0.49 0.00 1.00
B20 -0.80 -0.72 -0.43 -0.20 -0.62 -0.12 -0.72 1.00
HE 0.43 0.82 0.36 0.00 0.78 0.23 0.63 -0.66 1.00
Un 0.46 0.43 0.28 0.36 0.57 0.31 0.29 -0.56 0.37 1.00
DR -0.04 -0.23 -0.22 0.30 -0.09 0.16 -0.10 0.07 -0.16 -0.12 1.00
Mean 0.36 1400.88 112.58 4.13 0.53 6.70 5.77 11.22 4.00 4.75 101.67
S. Deviation 0.24 355.84 105.21 5.37 0.12 0.66 8.09 4.12 3.64 2.76 78.56
119
(ii) Overall Areas (combined Provinces)
In case of overall areas, most of the independent variables have low correlation
with BNGI. The results are shown in Table 5.2. If we consider the minimum criterion of
absolute correlation as r > [0.1]; then the following seven regressors can be considered
as significant: Rem, HCI, T2B, B20, HE, Un and DR. Interestingly, the per capita income
has low correlation with the dependent variable, but this feature can be ignored and YPc
cannot be excluded due to its importance and that all other significant variables are
highly correlated with it.
Table 5.2 Correlations, Means and Standard Deviations Overall Areas
Variables BNGI Ypc Spc Rem HCI HS T2B B20 HE Un DR
BNGI 1.00
Ypc 0.07 1.00
Spc -0.06 0.63 1.00
Rem 0.24 -0.14 -0.24 1.00
HCI 0.23 0.39 -0.07 0.50 1.00
HS 0.06 -0.13 -0.26 0.56 0.55 1.00
T2B 0.61 0.57 0.19 0.20 0.42 -0.03 1.00
B20 -0.70 -0.52 -0.17 -0.24 -0.37 0.01 -0.90 1.00
HE 0.10 0.60 0.08 0.03 0.54 0.23 0.34 -0.34 1.00
Un 0.23 0.12 0.03 0.38 0.48 0.31 0.31 -0.35 -0.01 1.00
DR -0.31 -0.54 -0.23 -0.18 -0.60 -0.10 -0.56 0.53 -0.56 -0.33 1.00
Mean 0.39 1291.53 88.28 4.63 0.50 6.63 3.92 11.19 2.88 3.87 95.97
S. Deviation 0.18 179.29 68.90 0.36 0.07 0.63 1.56 2.45 1.94 1.70 9.01
120
Next we look at the horizontal position, i.e. values in a row from left to right. It is
observed that Remittances and HCI are highly correlated ( 0.5r ) and one variable
ought to be deleted; same is case for HE and HCI pair; and for Un and HCI pair. On the
other hand, DR is highly correlated with HCI, T2B, B20 and HE and likewise B20 and
T2B are highly correlated, Keeping in view these considerations, if Rem, DR, HE and
either T2B or B20 are excluded, the problem of multi-collinearity is solved to a great
extent. The model will assume the final form to include YPc, HCI, B20, Un as significant
explanatory variable and look similar to equation 5.2 shown above.
(iii) Rural Areas (of the four Provinces)
The results are shown in Table 5.3. In the rural areas, all the variables with the
exception of T2B and B20 have low correlation with the dependent variable (BNGI).
Therefore a criterion of r > [0.13] is considered for the purpose. This pinpoints the
variables Spc, Rem, HCI, T2B, B20, and Un to be significant. For possible conflicts in
data, the per capita income shows lower correlation with the dependent variable.
However, this variable cannot be excluded for reasons explained above and further its
close associate, namely the per capita saving has some acceptable correlation with
BNGI.
Next we look at the cross section of the selected variables to see their mutual
correlation. The correlation between HCI and Rem is high [ 0.67r ], and between B20
and T2B is high as usual [ 0.86r ]. If we exclude the variables Rem, and T2B, the final
model may be „good fit‟ and should comprise the four regressors, namely: YPc instead
of Spc, HCI, B20, and Un. The model may be written as equation 5.2 shown above.
121
Next we look at the horizontal position, i.e. values in a row from left to right. It is
observed that Remittances and HCI are highly correlated ( 0.5r ) and one variable
ought to be deleted; same is case for HE and HCI pair; and for Un and HCI pair. On the
other hand, DR is highly correlated with HCI, T2B, B20 and HE and likewise B20 and
T2B are highly correlated, Keeping in view these considerations, if Rem, DR, HE and
either T2B or B20 are excluded, the problem of multi-collinearity is solved to a great
extent. The model will assume the final form to include YPc, HCI, B20, Un as significant
explanatory variable and look similar to equation 5.2 shown above.
Table 5.3 Correlations, Means and Standard Deviations Rural Areas (four provinces)
Variables BNGI Ypc Spc Rem HCI HS T2B B20 HE Un DR
BNGI 1.00
Ypc -0.02 1.00
Spc -0.16 0.80 1.00
Rem 0.23 0.08 -0.15 1.00
HCI 0.13 0.20 0.03 0.67 1.00
HS -0.10 -0.08 -0.19 0.56 0.70 1.00
T2B 0.60 0.50 0.20 0.32 0.27 -0.08 1.00
B20 -0.60 -0.49 -0.14 -0.34 -0.17 0.13 -0.86 1.00
HE -0.11 0.20 0.06 0.34 0.42 0.49 -0.06 -0.04 1.00
Un 0.28 0.18 0.01 0.46 0.50 0.33 0.35 -0.40 0.06 1.00
DR 0.10 -0.16 -0.23 0.34 0.18 0.25 0.27 -0.14 0.04 0.00 1.00
Mean 0.28 1118.91 62.69 4.92 0.44 6.57 2.61 13.51 1.34 3.43 116.50
S. Deviation 0.21 156.05 84.99 6.73 0.08 0.72 1.13 3.11 1.03 1.58 109.18
122
(iv) Urban Areas (of the four Provinces)
Finally, we repeat the exercise for the urban areas of the four provinces. The
results are shown in Table 5.4. In this case, most of the proposed variables have
relatively high correlation with dependent variable when compared to the rural areas. If
the limit of r>0.4 in absolute value is set the criterion, then Ypc, Rem, HCI, T2B, B20,
HE, Un and DR turn out to be significant so far as their relationship with BNGI is
concerned.
Table 5.4 Correlations, Means and Standard Deviations
Urban Areas (four provinces) Variables BNGI Ypc Spc Rem HCI HS T2B B20 HE Un DR
BNGI 1.00
Ypc 0.47 1.00
Spc 0.29 0.25 1.00
Rem 0.44 0.15 -0.02 1.00
HCI 0.45 0.54 -0.24 0.48 1.00
HS 0.12 -0.33 0.25 0.15 -0.06 1.00
T2B 0.73 0.65 0.25 0.17 0.44 -0.14 1.00
B20 -0.93 -0.57 -0.29 -0.45 -0.54 -0.16 -0.79 1.00
HE 0.44 0.66 0.01 0.16 0.63 0.04 0.58 -0.60 1.00
Un 0.41 0.06 0.10 0.76 0.31 0.24 0.12 -0.42 0.02 1.00
DR -0.66 -0.61 0.00 -0.51 -0.51 0.16 -0.62 0.72 -0.50 -0.40 1.00
Mean 0.44 1682.84 162.47 3.34 0.62 6.83 8.93 8.93 6.66 6.07 86.84
S. Deviation 0.25 263.13 100.27 3.41 0.07 0.58 10.51 3.72 3.36 3.05 9.79
However, when the mutual correlation among these variables is considered, we
observe that B20 and T2B are highly correlated as usual [ 0.79r ]. Likewise, HE has
123
high correlation with HCI [ 0.63r ] and B20 [r=-0.60], Un and Rem are highly
correlated [ 0.76r ], finally DR and B20 are highly correlated [ 0.72r ]. The high
correlation of some variables with per capita income can be ignore for plausible
reasons. Keeping in view these results, HE, Rem, T2B and DR can be excluded from
analysis. The model assumes the final form as under to include four explanatory
variables, Ypc, HCI, B20 and Un, and the equation for the purpose of regression
analysis will look similar to 5.2 shown above.
5.1.2 Static Panel Models
Consider a panel description of the relation between dependent and independent
variable given as it it itY X , where is vector of coefficients. The coefficients are
assumed to be constant, i.e. they do not vary over time and space. Under this
assumption, the panel specification becomes a static model.
Although the assumption of invariability of the coefficients for all cross-sections is
not reasonable due to heterogeneity among the cross-section, the model has the
capability to capture co-variation of the regressors and regressand. Furthermore, as the
sample size increases when we merge all the cross-sections, the degree of freedom
also increases, assuming the coefficients to remain unchanged. Therefore the model
provides precise estimates of the parameters. The model in equation 5.1 is estimated
using static panel model technique (OLS method) for four different groups of data, i.e.
Rural Areas, Urban Areas, Overall Areas, and aggregate Rural Urban Areas. The
results are discussed below:
124
(a) Rural Areas (of the four Provinces)
The regression results for the four rural areas are shown in Table 5.5.
Table 5.5 Static Panel Model for Rural Areas Dependent Variable BNGI
Coefficient Std.Error t-value t-prob
Ypc -0.001005*** 0.0002015 -4.99 0.000
Spc 0.00052*** 0.0001345 3.88 0.000
Rem 0.000312 0.003780 0.0825 0.935
HCI 0.2627 0.4473 0.587 0.560
HS -0.01829 0.03197 -0.572 0.570
T2B 0.1016*** 0.02102 4.84 0.000
B20 -0.0304** 0.01207 -2.52 0.016
HE 0.00708** 0.002911 2.43 0.019
Un 0.00124 0.01560 0.0794 0.937
DR -0.00034*** 9.068e-005 -3.79 0.000
Constant 0.54853*** 0.2355 6.57 0.000
Sigma 0.1485222 Sigma^2 0.02205884
R^2 0.5872018
RSS 0.90441231308 TSS 2.190931
No. of observations 52 No. of parameters 11
Using robust standard errors
AR(1) test: N(0,1) = 1.194 [0.232]
AR(2) test: N(0,1) = -0.2814 [0.778]
Note: Statistical significance is indicated by asterisk signs.
*** Significant at 1% level, ** Significant at 5% level, * Significant at 10% level.
_______________________________________________________ Calculations are based on Pc-give (Ox Metrics)
The above results have been obtained for the rural areas (balanced panel). The
objective is to arrive at the suitable specification. Per capita income (Ypc), per capita
savings (Spc), ratio of income of top to bottom 20% (T2B), share of income held by the
bottom 20 % (B20), Higher Education (HE), and Dependency Ratio (DR) appear to be
125
statistically significant using robust standard errors. The value of R2 indicates that model
is moderately fit. Since P-value of AR(1) and AR(2) tests are greater than 10 % (as well
as 1% and 5%), so we do not reject the null hypothesis of presence of 1st and 2nd order
autocorrelation, particularly between Ypc and Spc and between B20 and T2B.
Therefore, the variables Spc and T2B ought to be deleted from the model. The final
equation will again look like equation 5.2.
(b) Urban Areas (of the four Provinces)
Table 5.6 Static Panel Model for Urban Areas Dependent Variable BNGI
Coefficient Std.Error t-value t-prob
Ypc -9.47855e-005*** 2.962e-005 -3.20 0.003
Spc 0.0001104* 6.162e-005 1.79 0.081
Rem 0.00213 0.006736 0.316 0.753
HCI 0.102 0.1411 0.722 0.474
HS -0.027 0.01850 -1.50 0.142
T2B 0.00048 0.001046 0.464 0.645
B20 -0.0679*** 0.006084 -11.2 0.000
HE -0.0109*** 0.003807 -2.87 0.007
Un -0.00283 0.002990 -0.947 0.349
DR -0.00057 0.0008155 -0.706 0.484
Constant 1.43814*** 0.2828 5.09 0.000
Sigma 0.09468026 Sigma^2 0.008964352
R^2 0.8821278
RSS 0.3675384232 TSS 3.1181106731
No. of observations 52 No. of parameters 11
Using robust standard errors
AR(1) test: N(0,1) = 0.5297 [0.596]
AR(2) test: N(0,1) = -0.08453 [0.933]
___________________________________________ Calculations are based on Pc-give (Ox Metrics)
126
Table 5.6 shows results for the urban areas obtained, using the static panel
method. According to this regression, the per capita income (Ypc), per capita savings
(Spc), share of income held by the bottom 20 % (B20), and Higher Education (HE)
appear to be statistically significant. However, the income per capita and saving per
capita may be highly correlated as indicated by AR (1) and AR (2) tests. Similarly, the
human capital (HCI) may be better variable than higher education (HE). The high value
of R2 is indicative of good fit. It is therefore safe to consider the model given by equation
5.2 to be appropriate.
(c) Overall Areas (of the four Provinces)
Table 5.7 indicates the estimation results of overall areas through static panel. It
is clear from the table that per capita income (Ypc), per capita savings (Spc), ratio of
income of top to bottom 20% (T2B), and share of income held by the bottom 20 %
(B20), are significant at 1% level. By omitting the insignificant variables, there is no
danger of specification bias. However, Ypc and Spc may be highly correlated as also
T2B and B20. Therefore, it is safe to delete Spc and T2B to arrive at the appropriate
model, which may look like equation 5.2. The value of R2 (0.65) is indicative of
reasonably good fit.
127
Table 5.7 Static Panel Model for Overall Areas Dependent Variable BNGI
Coefficient Std.Error t-value t-prob
Ypc -0.00071*** 9.221e-005 -7.73 0.000
Spc 0.00049*** 0.0001492 3.29 0.002
Rem -0.004 0.004079 -0.981 0.332
HCI 0.623 0.5089 1.22 0.228
HS -0.0013 0.02001 -0.0630 0.950
T2B 0.0153*** 0.005011 3.05 0.004
B20 -0.0654*** 0.007193 -9.09 0.000
HE 0.00232 0.01247 0.186 0.854
Un -0.0127 0.009939 -1.29 0.206
DR -0.000164 0.001513 -0.108 0.914
Constant 1.68827*** 0.1194 14.1 0.000 Sigma 0.1188651 Sigma^2 0.01412891
R^2 0.6504891
RSS 0.57928543997 TSS 1.6574174423
No. of observations 52 No. of parameters 11
Using robust standard errors
AR(1) test: N(0,1) = 0.7601 [0.447]
AR(2) test: N(0,1) = -0.9316 [0.352]
--------------------------------------------------------------------------
Calculations are based on Pc-give (Ox Metrics)
(d) Aggregate Rural and Urban Areas (of the four Provinces)
Table 5.8 shows the results for aggregate rural and urban areas of the four
provinces. Interestingly only three variables, namely the per capita income (Ypc), ratio
of income of top to bottom 20% (T2B), and share of income held by the bottom 20 %
(B20), are significant. Due to high correlation of B20 and T2B, the latter may be dropped
from the analysis. Instead, we will prefer to adopt the model as per equation 5.2, which
128
is supported by the tests mentioned above. Again, in this case, the high value of R2
(0.718) indicates that the model is good fit.
Table 5.8 Static Panel Model for Aggregate Rural Urban Areas Dependent Variable BNGI
Coefficient Std.Error t-value t-prob
Ypc -0.000303** 0.0001161 -2.61 0.011
Spc 2.99004e-005 0.0001695 0.176 0.860
Rem -0.000689065 0.003415 -0.202 0.841
HCI 0.332325 0.2553 1.30 0.196
HS -0.0242700 0.01992 -1.22 0.226
T2B 0.007185*** 0.002396 3.00 0.003
B20 -0.05055*** 0.005299 -9.54 0.000
HE -0.00433 0.003510 -1.23 0.220
Un 0.00395 0.005955 0.663 0.509
DR -8.30044e-005 0.0001200 -0.692 0.491
Constant 1.30175 *** 0.1689 7.71 0.000
Sigma 0.1339398 Sigma^2 0.01793988
R^2 0.7181438
RSS 1.6684085454 TSS 5.9193598365
No. of observations 104 No. of parameters 11
Using robust standard errors
AR(1) test: N(0,1) = 0.9214 [0.357]
AR(2) test: N(0,1) = -0.8521 [0.394]
__________________________________________
Calculations are based on Pc-give (Ox Metrics)
5.1.3 Impulse Saturation
The Impulse Saturation is the most advanced technique developed by David F.
Hendry and others at the London School of Economics in the year 2008. This technique
takes into account multiple criteria simultaneously, including the significance of
regressors, model adequacy, and the forecast performance. The estimation starts with a
129
General Unrestricted Model (GUM) and then the variables are dropped one after the
other, provided they do not adversely affect the three criteria mentioned above. This
technique can be applied to any model using Pc-Give software. The present study
employs the above software proposed by Doornik (2009). The print-out is quite lengthy
and we reproduce only the final equation, which is also known as auto-selected model.
Other initial calculations are omitted for the precision purpose. The results of the model
for rural, urban, overall areas and aggregate rural and urban areas are given below in
Tables 5.9, 5.10, 5.11 and 5.12. We discuss the results briefly.
(1) Rural Areas (auto-Selected Model)
Table 5.9 Rural Areas (Auto Selected Model)
Dependent Variable BNGI Coefficient Std. Error t-value t-prob Partial R^2
Constant 1.1523 0.3092 3.73 0.0005 0.22
Ypc -0.0006 0.0002 -4.07 0.0002 0.26
T2B 0.0826 0.0360 2.30 0.0261 0.10
B20 -0.0294 0.0129 -2.28 0.0268 0.10
Sigma 0.144455 RSS 1.00162744
R^2 0.54283 F(3,48) = 19 [0.000]**
Log-likelihood 28.9053 DW 1.57
No. of observations 52 No. of parameters 4
Mean(BNGI) 0.2825 Var(BNGI) 0.0421333
-------------------------------------------------------------------------------------------
Calculations are based on Pc-give (Ox Metrics)
The results reported in Table 5.9 indicate that out of the ten explanatory
variables, only three are significant, i.e. the per capita income (Ypc), ratio of income of
top to bottom 20% (T2B), share of income held by the bottom 20 % (B20). The program
has omitted all other variables automatically. The value of R2 shows average fit whereas
130
Durbin Watson statistics shows that null hypothesis can neither be accepted nor
rejected. Small value of variance of BNGI however suggests that it is consistent.
131
(2) Urban Areas (auto-Selected Model)
The results for urban areas for same number of observations are reported in
Table 5.10. Interestingly only two variables are reported to be significant and pass all
the tests of model selection, namely the share of income held by the bottom 20 %
(B20), Higher Education (HE). The value of R2 indicates that the model is very good fit
while DW value (1.93) suggests that null hypothesis of „No Autocorrelation of first order‟
cannot be rejected in the data. As stated earlier, we may keep the per capita income for
its significance on theoretical basis. Likewise, we may substitute the human capital
index (HCI) for higher education, being more comprehensive.
Table 5.10 Urban Areas (Auto Selected Model) Dependent Variable BNGI Coefficient Std.Error t-value t-prob Part.R^2
Constant 1.1338 0.0616 18.40 0.0000 0.87
B20 -0.0685 0.0041 -16.60 0.0000 0.85
HE -0.0130 0.0046 -2.83 0.0068 0.14
Sigma 0.0880839 RSS 0.380179475
R^2 0.878074 F(2,49) = 176.4 [0.000]**
Log-likelihood 54.0924 DW 1.93
No. of observations 52 No. of parameters 3
Mean (BNGI) 0.435712 Var (BNGI) 0.0599637
-----------------------------------------------------------------------------------------------------
Calculations are based on Pc-give (Ox Metrics)
(3) Overall Areas (auto-Selected Model)
When overall areas are taken into account and the data confronted to the test for
model selection, the final reports obtained are shown in Table 5.11. As evident, the per
capita income (Ypc) and share of income held by the bottom 20 % (B20) come out to be
132
significant. All other measures R2, DW, Var (BNGI), F-test are reasonably satisfactory to
accept the model.
Table 5.11 Overall Areas (Auto Selected Model) Dependent Variable BNGI Coefficient Std.Error t-value t-prob Part.R^2
Constant 1.6443 0.1939 8.48 0.0000 0.59
Ypc -0.0004 0.0001 -3.91 0.0003 0.24
B20 -0.0673 0.0076 -8.81 0.0000 0.61
Sigma 0.114144 RSS 0.638415411
R^2 0.614813 F(2,49) = 39.11 [0.000]**
Log-likelihood 40.6155 DW 1.77
No. of observations 52 No. of parameters 3
Mean (BNGI) 0.364327 Var (BNGI) 0.0318734
----------------------------------------------------------------------------------------------------------
Calculations are based on Pc-give (Ox Metrics)
(4) Aggregate Rural-Urban Areas (auto-Selected Model)
Table 5.12 Aggregate Rural and Urban Areas
Dependent Variable BNGI Coefficient Std.Error t-value t-prob Part.R^2
Constant 1.2373 0.1179 10.50 0.0000 0.52
Ypc -0.0002 0.0001 -4.41 0.0000 0.16
T2B 0.0065 0.0024 2.71 0.0078 0.07
B20 -0.0521 0.0052 -10.00 0.0000 0.50
Sigma 0.131443 RSS 1.72772233
R^2 0.708123 F(3,100) = 80.87 [0.000]**
Log-likelihood 65.5049 DW 1.75
No. of observations 104 No. of parameters 4
Mean (BNGI) 0.359106 Var (BNGI) 0.0569169
-------------------------------------------------------------------------------------------------------------
Calculations are based on Pc-give (Ox Metrics)
The results for aggregate rural and urban areas are presented in Table 5.12.
Here the per capita income (Ypc), ratio of income of top to bottom 20% (T2B), and
133
share of income held by the bottom 20 % (B20) appear to be statistically significant.
Value of R2 indicates that model is good fit. DW statistic shows that we cannot accept
the hypothesis of autocorrelation to be present in the data.
The results for different regions of Pakistan using the above mentioned
techniques suggest different forms of the model. In most of the cases, the income per
capita (Ypc), the human capital index (HCI), the share of income held by the poor at the
bottom 20 % of population (B20) and unemployment rate (Un) appear to be significant.
The appropriate model will again look like that given in equation 5.2, reproduced below.
0 1 2 3 420i pc i i i i iBNGI Y HCI B Un V (5.2)
The objective of the present study is to empirically analyze the determinants of
basic needs fulfillment (BNF) using the data from different regions of Pakistan. The
study empirically reviews the basic needs approach to development and poverty
reduction strategies. We have employed different estimation techniques in a quest to
arrive at the appropriate and plausible determinants of basic needs gap index. There is
also a descriptive analysis of the different regions of Pakistan analyzing the present and
past state of basic need fulfillment and its linkage with the poverty. The analysis is
primarily focused on the comparison among various regions of Pakistan (rural as well as
urban). Based on the data set, and model restrictions, suitable econometric techniques
are used and compared to obtain the reliable estimates in the next sections.
134
5.2 Empirical Model
The relationship between the dependent and independent variables can be
summarized via panel representation, reproduced below:
1
K
it ki kit it
k
Y X
where
1,2,3....,i N and 1,2,3....,t T
In the above relation, KX is kth regressor and the subscript „ i „ denotes the region
concerned and „ t „ denotes variation overtime. The study will employ two econometric
techniques to estimate the model given in (5.2). The first one is Region Specific Least
Square (RSLS) method that provides the initial estimation. The second one is the
application of the Bayesian (empirical Bayes) procedure that utilizes the OLS estimates
as priors. This technique has two alternative versions, the first is due to Carrington &
Zaman (1994) and the second is due to Hsiao and Pesaran (2004). The rationale for
using these econometric techniques, the estimation procedure and simplifying
assumptions are explained in the next section.
5.2.1 Classical Bayes vs. Ordinary Least Square (OLS)
In case of ordinary least square (OLS), prior information about the parameters is
ignored and the parameters are assumed to be fixed. In contrast, the Bayes method
includes the prior knowledge about the parameters and has an improvement over the
OLS estimators; it further suggests a subjective explanation of statistics as opposed to
an objective analysis. For example, if we are interested in estimation of consumption
function, we may follow the least squares as well as the Bayes procedure. In case of
135
classical least squares, the parameters of the function are estimated and they are
assumed to possess certain characteristics, called the BLUE properties (best linear
unbiased estimator). The model assumes that the true population parameters are
constant or not varying. However, the consumption habits in the economy are random
and likely to change over time. Moreover, the parameter estimates should have some
desirable asymptotic properties, i.e. if sufficiently large data are available the estimates
tend to converge to the true value of the population parameter.
In the classical (OLS) case, the emphasis is on the estimator itself and the
statistics that describe it, whereas in Bayesian analysis, the explanation of an estimator
is fairly different. Contrary to making a point estimate, the Bayesian method formulates
a posterior density function for the data, which is different from a sampling distribution.
This can be understood with reference to a prior confidence about what one
believed. It is normally discussed as the odds a researcher would give when taking bets
on the true value of data. Therefore, one needs to identify the initial degree of belief
while using empirical evidence as a means of changing that belief. Thus, the Bayesian
approach, draws upon a prior density function as well as a posterior density function,
which is a stark improvement over the classical (OLS) approach.
According to Greene (2004), if the empirical results conflict with the theory, such
results are not accepted by the classical approach and these are liable to collapse in
order to uphold the theory. But in case of the Bayesian approach, theory can be
reformulated. The existing evidence is assembled and compared with the theory, beliefs
are formulated and based upon the existing evidence, then further evidence is collected
136
and previous beliefs are compared with the new evidence. In this way, the beliefs
regarding the theory are revised.
Berger (1985) mentioned the following advantages of Bayesian analysis:
Firstly, contrary to classical estimation, the Bayes method assumes that the
parameters are random with prior density. Due to this characteristic, Bayesian
estimation is considered suitable for panel data.
Second, the Bayesian method offers a natural way of comparing and contrasting
prior belief with the available evidence (data). The parameter estimates obtained
through the classical method in case of panel data can be used as prior. That is why
this is suitable technique for panel data models.
Third, the Bayesian estimates are more precise when compared with the
classical estimates i.e. the Bayesian estimates have smaller standard errors, and hence
the results are more reliable.
Fourth, the Bayesian procedure gives precise and reliable results even in case of
small samples whereas the classical procedure, estimates are consistent only
asymptotically, i.e. in large samples.
5.2.2 Bayesian Estimation Procedure
Suppose the model is given below in the compact format, where Y denotes the
dependent variable, the set of independent variables is given by the matrix X (of order
nxk), β is the column vector of the parameters and ε is the vector of random error.
137
Y= X
The maximum likelihood (ML) estimator of β is given by
1ˆ ( )X X X Y
Under the standard assumptions, ̂ has the following density
2 1ˆ ˆ~ ( , (X X) ) where the variance is given by 2
ˆ ˆ( ) ( )ˆ
( )
Y Y
T K
The Bayesian estimation procedure assumes the population parameter to be
randomly described by the density function ~ ( , )N where and are called
priors since they represent our prior knowledge about the parameters.
The posterior estimate is then the weighted average of the prior and data mean and is
given by
2 1 1 2 1ˆ ˆ ˆ( ) ( (X X) ) ( (X X) , )B E (5.3)
and the posterior variance is given by
1 1
2
1ˆ( ) ( )Var X X
2
1X X
is the inverse of the variance of the data density, called the precision of data
density, and 1 is the precision of the priors. The variance of posterior density is given
by: 1 1
2
1ˆ( ) ( )BayesVar X X
(5.4)
138
so that posterior precision is 1 1
2
1( )X X
This is the sum of the prior and the data precision. Therefore the Bayesian estimator is
more precise than the data and the prior.
5.2.3 Classical Bayes vs. Empirical Bayes
For the most part of the research work, the economists have generally avoided to
use the Bayesian methods. The reasons for this attitude are not because they object to
the underlying philosophical nature of subjectivist probability; instead, there are practical
reasons. Zaman, (1996) noted three major problems with the use of Bayesian
estimators in practice given below:
First, the Bayesian models may have unbounded risk, depending on choice of
priors. If prior is precise enough, the improvement over maximum likelihood
estimator is substantial. If the prior is less precise, then the improvement over ML
estimator is very small and therefore fruitless.
Second, there is the problem of choice of hyper parameters. The classical
Bayesian procedure for the choice of hyper parameters is arbitrary and there are
no specific rules for choosing priors.
Third, sometimes the conflict between data and prior creates problems for
investigators.
Empirical Bayes method is especially formulated to avoid these three difficulties
by making the selection of the priors only after looking at the data, and fixing the values
of the priors according to those of the data. Empirical Bayes technique gives more
139
precise results particularly when sample is small and where the OLS method is
imprecise (Carrington & Zaman, 1994). Due to this imprecision of OLS estimation, the
present study is also using empirical Bayes to substantiate the OLS estimation.
Bayesian estimation takes advantage of the prior knowledge obtained through OLS or
M.L. estimation. Empirical Bayes uses aggregate information from the data as prior and
hence gives precise estimates. Techniques for application of Bayes estimators are
given in Zaman (1996) and Geweke (2005).
5.3 Estimation Approaches
As discussed above, the present study uses three different approaches to find
the estimates. These approaches are as follows:
Ordinary least square (OLS) [Region Specific Least Square].
Empirical Bayes following Hsiao and Pesaran (2004)
Empirical Bayes following Carrington and Zaman (1996) and
The imprecision problem of OLS estimates, if any, can be overcome by employing the
empirical Bayes. We discuss briefly the two versions of the empirical Bayes
methodology in the following lines.
5.3.1 Empirical Bayes due to Hsiao and Pesaran (2004)
As the model (equation5.2) consumes a lot of degrees of freedom and lacks
explanatory power, therefore this type of model can be estimated by imposing certain
assumptions about the parameters to get the predictive power.
Consider the following model:
140
' '
it it it i i ittY Z W
where
1,2,3....,i N and 1,2,3....,t T (5.5)
itZ and itW are vectors of exogenous variables with dimension ( &l p ), where
1i
n is the average of all individual coefficients, and i i is the deviation
of individual specific effects from the common effects. i is assumed to be random with
mean 0 and variance .
In practical situations, '
itZ and '
itW may be equal to the vector of regressors '
itX ; in
which case the model reduces to
' ( )it it it itY X (5.6)
This is similar to assuming that it it
with
We further assume that the parameter value doesn‟t change with time, in which case we
get it i
, and i is independently distributed multivariate normal with (0, )i N
The model finally reduces to
itit it itY X , (5.7)
which looks similar to equation 5.1 that we wanted to estimate. This formulation is
useful to decompose the coefficients into general and region specific effects. Under the
simplifying assumptions discussed above, the final empirical model can be formulated
as:
Y Z W U (5.8)
141
where 1
1
2
NT
N
y
yy
y
1
1
2
T
i
i
i
iT
y
yy
y
, 1
1
2
NT
N
U
UU
U
,
1
2
1
i
i
iT
iT
u
uU
u
,
1
2
NT l
N
Z
ZZ
Z
,
1
2
i
i
iT l
iT
Z
ZZ
Z
1
2
0 0
0NT Np
N
W
W W
W
'
1
'
2
'
i
i
iT p
iT
w
wW
w
and
1
2
1Np
N
Is the proportion of coefficients of regressor that doesn‟t change with regions and i
contains information on region specific effects.
142
Following Hsiao and Pesaran (2004), it is assumed that:
Prior distribution of and i are independent. That is ( , ) ( ). ( )P P P
The prior for is flat, which states that we have no information about , i.e.
( )P constant.
Prior Distribution of is known i.e. (0, )NN I
Under these assumptions, and given and , we have ( , )Y N Z W C ,
where cov( )C . The empirical Bayes estimates the prior parameters from the
marginal distribution, which in turn is derived by Hsiao and Pesaran (2004) as under:
( , ( ) ))NY N Z C W I W (5.9)
The matrix Z contains the exogenous regressors as defined in equation 5.8. Therefore,
can be obtained as least square estimate of by regressing Y on Z. However, we
should have estimates of C and to compute posterior estimates of parameters. Now
if the classical Bayes procedure is followed, value of and C is computed as modal
value of and C for the arbitrary choice of α and . The iteration starts by assuming
some arbitrary value of and , for any given value of and , one can compute
value of C and using many such values of and .
The mode of the values of C and are chosen as the hyper parameters.
143
The empirical Bayes procedure differs from the classical Bayes in that the hyper
parameters are also estimated from data.
Therefore, the variance covariance matrix of i given by can be estimated as
1( )( )i i
N
(5.10)
Now to estimate C we have to find out estimated covariance matrix of errors for all cross
sections. Since errors are assumed iid, the covariance matrix takes the form;
2
1
2
2
0 0
0 0
0 0
T
T
I
C I
Where 2
/i i T K and is given in equation (5.7).
when we have and C the posterior estimates of individual effects and their respective
variance covariance matrix is given by,
1 1 1 1 1 1 1 1 1 1 1 1{ [ ( ) ] ( )} { [ ( ) ] }NW C C Z Z C Z Z C W I W C C Z Z C Z Z C Y
(5.11)
and 1 1 1 1 1 1 1{ [ ( ) ] ( )}ND W C C Z Z C Z Z C W I
(5.12)
5.3.2 Empirical Model Following Carrington and Zaman
According to Carrington and Zaman (1994), the relationship between dependent
and independent variables can be summarized in m regression models of the form,
it it i iY X
1,2,...,i m (5.13)
144
where itX is the matrix of regressors for region i , itY is the dependent variable, and
i is the vector of coefficients for the region, which is assumed to be randomly
distributed with some mean and variance: ~ ( , )i N where is vector of the
common effects for all the regions.
The OLS estimate of equation (5.13) is then 2 2 ' 1, , ( )OLS
i i i i i i iN X X
where
' 1 '( ) .OLS
i i i i iX X X Y (5.14)
Empirical Bayes assumes that the true parameter values for the individual
regions are inter related and that i has a normal prior distribution of the form
[ , ] ( , ).i N (5.15)
The exchangeability condition assumes some centralized point around which the
individual region parameters are likely to be normally distributed. This provides a
justification for cross regional analysis within a country instead of cross country
analysis, where exchangeability condition is weak.
Having exchangeability assumption, we specify the hyper parameters and
and find empirical Bayes estimates 'i s , where the Bayesian estimator is given by:
1 2 ' 1( ),Bayes OLS
i i i i i iD X X (5.16)
where 2 ' 1
i i i iD X X (5.17)
145
This is the weighted average of the OLS estimate, and the assumed prior mean, where
the weights are the estimated variances of the OLS estimators and the assumed prior
variance.
Empirical Bayes (EB) allows and to be estimated directly from the provinces
distribution of the OLS parameters and the estimate of is found by,
1
' '
2 21 1
1 1T T
i i i i
i ii i
X X X Y
(5.18)
which is weighted average of the province specific OLS estimates, where the weights
are inversely related to the parameters estimated variance.
To find , we follow Blattberg and George (1991) and Carrington and Zaman
(1994), and restrict the off diagonal elements of to be zero so that
1 2 7( , ,..., ),diag which assumes no prior covariance among the coefficients.
2 2
1
1( )
1
T
i ij j i ij
i
aT
(5.19)
We then re-estimate the 'i s with (5.18, 5.19) and re-estimate each element of with
1
' 1 ' 1
2 21 1
1 1( ( ) ) ( ( ) )
T T
j i i j i i i j i ij
i ii i
X X X Y
(5.20)
Having solutions of (5.16), (5.19) and (5.20), the estimated variance of the posterior
distribution of the
'i s are obtained as:
146
1
11 2 ' 1var( ) ( )EB
i i i iX X
(5.20)
This is smaller than the variance of the OLS estimator. This precision is due to the
increased information incorporated into the model.
Here we assess and determine impact of various factors on BNGI across
different regions (rural/urban) of Pakistan. Various regressors like per capita income
Ypc, human capital index HCI, ratio of income of richest 20% to bottom 20% T2B, share
of income held by bottom 20% B20, higher education HE and unemployment Un are
included to see their impact on BNGI. However the final model includes only four
regressors i.e. Ypc, HCI, B20, and Un. The results are reported in the next chapter,
employing all the techniques of analysis.
147
CHAPTER 6
EMPIRICAL RESULTS AND ANALYSIS
This chapter presents results obtained by confronting the data to model
discussed in chapter 4 and 5 via the estimation techniques namely the OLS and
empirical Bayes. The model is reproduced for ready reference (Equation 5.2):
itititititpcit UnBHCIYBNGI 432,10 20
As already discussed, most of the data is drawn from various issues/editions of
the HIES/ PSLM/ Economic Survey/ Labour Force Survey and other official documents.
However, some of the composite variables, particularly BNGI and HCI, have been
constructed. The subscript ‘i’ stands for the region concerned and ‘t’ for time.
The results are presented under four different categories, i.e. (1) rural areas, (2)
urban areas, (3) overall areas province-wise, and (4) aggregate rural-urban areas. The
last category (aggregate rural-urban areas) is included for the advantage of having the
maximum number of observations, i.e. 104. Moreover, this category provides different
priors and different estimates for application of empirical Bayes techniques. This
difference also indicates that common priors cannot be taken in the estimation of rural,
urban and overall regions.
Apparently, the three approaches to estimation are based on similar
assumptions, but methodologically one is different from the other so far as precision of
results is concerned. OLS is widely used and it is appropriate for large samples.
However, it may lead to imprecise results particularly when the sample size is small.
148
Owing to these shortcomings in the least-squares method, the present study also
employs empirical Bayes approach that generates more precise estimates.
Another obvious reason to estimate and compare the results obtained through
the OLS with empirical Bayes is that the latter satisfies the parsimony criterion required
for model selection. This criterion states that simple explanation should always be
preferred to the complex one. In econometrics literature, if one obtains similar values of
R-Square in two regression models, the model with fewer explanatory variables needs
to be preferred. A practical advantage of parsimony is that one is unlikely to run up
against the degree of freedom problem. The number of observations in time series is
rarely large, so the lack of parsimony frequently leads to imprecise estimates. Still
another advantage of empirical Bayesian technique is that it can explain the results of
other models, specifically the Least Squares method.
The credibility and usefulness of any study can be judged by the major findings,
which are primarily based on the empirics. By adopting suitable methodology and
reaching at some meaningful results, one can assess the economic condition of the
country/region concerned. The economy of Pakistan is agro-based and majority of the
population is rural. Therefore, the rural sector should be given more importance. At the
outset, the empirical findings of the rural regions of Pakistan are discussed. We have
found the results of rural and urban areas to be more informative and meaningful when
taken separately as compared to the results of overall and aggregate rural-urban case.
We may expect the outcome or the impact of explanatory variables on the
dependent variable on the basis of economic theory. The dependent variable (BNGI)
149
measures the extent or depth of poverty. Naturally, the level of income per capita should
affect poverty negatively. Similarly, the share of income held by the bottom 20% of the
population should be negatively related to the dependent variable, i.e. the higher this
share, the lower will be poverty. On the other hand, the rate of unemployment should
have a positive impact on the extent of poverty, i.e. the higher the rate of
unemployment, deeper is the poverty. How the volume of human capital (education
level and health status) affect the dependent variable is a little bit complicated
phenomenon. If higher human capital leads to higher job opportunities/employment and
income generation, then the poverty level should go down. All these theoretical
predictions are supported only by the results obtained from urban Punjab.
The chapter consists of five sections. Section 6.1 is about results of rural areas
and section 6.2 is for the urban areas; whereas overall provinces are discussed in 6.3.
Section 6.4 presents the results of rural and urban areas by using the priors obtained
from the aggregate data i.e. eight rural-urban regions. Section 6.5 is devoted to
sensitivity analysis.
6.1 Rural Areas
As explained above, we analyze the data of rural areas of the four provinces
separately. So far as Bayesian estimation is concerned, we use the prior information
obtained from the data of all four rural regions in each case.
6.1 (a) Rural Punjab
We begin our analysis with the results obtained for rural Punjab, which are
shown in Table 6.1(a). First column shows the results of ordinary least squares (OLS).
The second column displays the results of empirical Bayes technique following Hsiao
150
and Pesaran (HP). The third column reports the results obtained through empirical
Bayes technique as suggested by Carrington and Zaman (CZ). The values of R-square
i.e. 0.82 in OLS, 0.75 in HP, and 0.80 in CZ are reasonably high reflecting the
appropriateness of our model.
The coefficient of per capita income (Ypc) is highly significant for both OLS and
CZ but it is insignificant in the case of HP. In column one, OLS estimates imply that a
unit increase in per capita income will result in decreasing the basic needs gap index by
0.001 units. The estimated coefficients of human capital index (HCI) for OLS, HP, and
CZ are statistically significant at different levels and indicate positive relationship with
BNGI for rural Punjab, which goes against the common wisdom. The share of income
held by bottom 20 percent (B20) shows a negative and significant impact on BNGI.
This result supports the theoretical perception. The results on unemployment (Un) show
that this factor does not carry any significant relationship with BNGI1. The negative sign
is contradictory to the conventional economic theory; however the reason behind this
might be the disguised employment in rural areas or some problems in data. Haider
(2006) is also skeptical for the data on unemployment in Pakistan.
1 Although it is significant at 10 % in OLS model (p=0.0658)
151
Table 6.1(a)
OLS and Empirical Bayes Estimates for Rural Punjab
Dependent Variable: BNGI
Least Squares Empirical Bayes HP Empirical Bayes CZ
CONSTANT 1.66** 1.09 1.49***
standard error 0.56 0.6693 0.3748
T- value 2.97 1.63 3.99
P- value 0.018 0.141 0.004
Ypc -0.001** -0.0005 -0.0008**
standard error 0.00 0.0004 0.0002
T- value -3.32 -1.35 -3.35
P- value 0.01050 0.2149 0.01007
HCI 2.7041*** 1.6205* 2.0139**
standard error 0.80 0.8159 0.64
T- value 3.38 1.99 3.13
P- value 0.0096 0.0822 0.0140
B20 -0.075*** -0.0624** -0.0712***
standard error 0.02 0.0238 0.015
T- value -3.52 -2.62 -4.57
P- value 0.0078 0.0306 0.0018
Un -0.082* -0.0290 -0.053
standard error 0.04 0.0492 0.03205
T- value -2.13 -0.60 -1.68
P- value 0.0658 0.5671 0.1307
R Square 0.82 0.75 0.80
Note: Statistical significance is indicated by asterisk sign.
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
6.1 (b) Rural Sindh
The estimates for rural Sindh are reported in Table 6.1(b). Conventionally, per
capita income is considered as an important variable that can reduce basic needs gap
index. The coefficient of this variable (Ypc) in rural Sindh, as expected, has negative
152
relationship with BNGI. Here (Ypc) is highly significant in case of OLS and CZ; whereas,
it is significant at 10 percent level in (HP).
Table 6.1(b)
OLS and Empirical Bayes Estimates for Rural Sindh
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 2.1*** 1.598** 1.84***
standard error 0.41 0.6290 0.3159
T- value 5.15 2.54 5.83
P- value 0.0008 0.03459 0.00030
Ypc -0.001*** -0.0007* -0.00096***
standard error 0.00 0.0004 0.00026
T- value -3.84 -1.89 -3.67
P- value 0.0049 0.09478 0.00629
HCI 1.44* 0.5683 1.319*
standard error 0.73 0.7889 0.60228
T- value 1.96 0.72 2.19
P- value 0.08510 0.49183 0.05977
B20 -0.058*** -0.0566** -0.053***
standard error 0.01 0.0214 0.01238
T- value -4.44 -2.65 -4.28
P- value 0.0022 0.02948 0.00267
Un -0.19* -0.0448 -0.169*
standard error 0.09 0.0574 0.08312
T- value -2.24 -0.78 -2.04
P- value 0.0550 0.45720 0.07516
R Square 0.81 0.65 0.81
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
The Human capital index (HCI) is positively correlated with BNGI in rural Sindh.
This implies that with an improvement in human capital index, BNGI widens. Again, this
result goes against the common wisdom. In other words, as education level and health
of people improves, the gap between the rich and the poor increases and the poor fail to
153
escape poverty line in rural Sindh. One plausible reason of such relationship might be
that people affording higher level of human capital belong to the middle and upper
class. In other words, health and education are not much important for the poor people
in Sindh or they have rare access to such facilities. The impact of the share of income
held by bottom 20% (B20) is negative with BNGI and the variable is highly significant.
The basic needs gap index is positively related by unemployment level in rural Sindh as
also observed in the case of rural Punjab.
6.1(c) Rural KPK (Khyber Pakhtunkhwa)
The OLS and empirical Bayes estimates for rural KPK are illustrated in Table
6.1(c). The results for Ypc are somewhat poor. Though the algebraic signs associated
with this variable are according to the prior expectations but the variable is insignificant
in reducing the poverty gap in rural KPK.
Contrary to the case of rural Punjab and rural Sindh, HCI bears the expected
negative sign in case of KPK, but it is statistically insignificant. The share of income held
by bottom 20% is significant for OLS and HP estimations. The unemployment rate has
positive relationship with BNGI (except in case of CZ) but is statistically insignificant.
The value of R-square signals that model is poorly fitted.
To summarize the results, all the variables carry the expected signs but in most
cases, they are insignificant so far their impact on the dependent variable is concerned.
The reason for this outcome may be problems in data or the vicious circle of poverty
which is very severe in case of KPT and Baluchistan.
154
Table 6.1(c)
OLS and Empirical Bayes Estimates for Rural KPK
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 2.79** 1.94** 1.58***
standard error 0.94 0.7524 0.434
T- value 2.99 2.59 3.65
P- value 0.017 0.032 0.006
Ypc -0.00102* -0.0006 -0.00058
standard error 0.00 0.0004 0.00031
T- value -2.10 -1.44 -1.82
P- value 0.068 0.188 0.10700
HCI -1.44 -1.169 -0.416
standard error 0.79 0.8021 0.54252
T- value -1.83 -1.46 -0.77
P- value 0.104 0.182 0.46421
B20 -0.06* -0.049* -0.027
standard error 0.03 0.0258 0.0222
T- value -1.87 -1.90 -1.24
P- value 0.09825 0.09428 0.249
Un 0.021 0.0333 -0.0222
standard error 0.05 0.0520 0.045
T- value 0.40 0.64 -0.49
P- value 0.701 0.539 0.634
R Square 0.51 0.35 0.39
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
6.1(d) Rural Balochistan
The estimated results for rural Balochistan are shown in Table 6.1(d). These
reflect similar picture as shown for rural Punjab and rural Sindh with the exception of
positive sign associated with unemployment. Although it is statistically insignificant but
155
the sign is correct it is in accordance with the conventional theory. The value of R-
square is smaller for rural Balochistan as compared to rural Punjab and rural Sindh.
Table 6.1(d)
OLS and Empirical Bayes Estimates for Rural Balochistan
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 1.93*** 1.48* 1.65***
standard error 0.50 0.6490 0.35
T- value 3.90 2.28 4.67
P- value 0.00450 0.05205 0.00160
Ypc -0.001** -0.0006 -0.0008***
standard error 0.00 0.0004 0.00023
T- value -3.22 -1.66 -3.53
P- value 0.01230 0.13513 0.00768
HCI 0.018 -0.1050 0.292
standard error 0.69 0.7941 0.57876
T- value 0.03 -0.13 0.50
P- value 0.97891 0.89747 0.62740
B20 -0.047** -0.044* -0.045***
standard error 0.02 0.0219 0.01333
T- value -3.07 -2.01 -3.40
P- value 0.01541 0.07955 0.00931
Un 0.032 0.040 0.027
standard error 0.04 0.0489 0.03283
T- value 0.92 0.82 0.81
P- value 0.38406 0.43543 0.44054
R Square 0.72 0.63 0.71
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
156
6.2 Urban Areas
The results obtained by OLS and empirical Bayes estimation for urban areas of
Pakistan scattered through the four provinces are shown in Table 6.2(a) to Table 6.2(d).
As explained above, we use the the prior information obtained from the data of all four
urban regions in the Bayesian estimation.
6.2 (a) Urban Punjab
Table 6.2(a) shows results for urban Punjab.
If we compare results for urban and rural Punjab, we come to conclude that all
the coefficients carry the correct signs but only one variable (B20) is statistically
significant. Ypc bears the correct sign for both areas, but it is insignificant in case of
urban area. However, HCI is negatively related to BNGI in urban Punjab, which is
according to expectations. This variable is statistically significant for empirical Bayes
(CZ). The share of income held by bottom 20% tells the same story for both rural and
urban areas of Punjab, which implies that the nature and impact of income distribution is
similar. Unemployment level in urban areas is positively related with the basic needs
gap, which is in accordance with the prior expectation. A high value of R- square signals
that model is good fit.
6.2 (b) Urban Sindh
Findings regarding the relationship of Ypc, HCI, B20, and Un with BNGI for urban
Sindh are given in Table 6.2(b). The coefficient on Ypc bears negative sign and is
statistically insignificant. The coefficient on HCI is negative for all the methods
employed, as compared to the positive values of rural Sindh and is significant in CZ.
157
The impact of unemployment is positive in case of OLS and empirical Bayes (CZ),
whereas it is negative for empirical Bayes (HP). These contradictory results pose some
questions about the quality of data about unemployment in Pakistan. The estimated
results regarding the share of income held by bottom 20%, presents similar trends for
both rural and urban areas.
Table 6.2(a)
OLS and Empirical Bayes Estimates for Urban Punjab
Dependent Variable: BNGI
OLS
Empirical Bayes HP
Empirical Bayes
CZ
CONSTANT 1.59*** 1.41** 1.33***
standard error 0.3890 0.4299 0.1783
T- value 4.10 3.28 7.51
P- value 0.0030 0.0110 0.0001
Ypc -0.0002 -0.00012 -0.0001
standard error 0.0002 0.0002 0.0001
T- value -0.90 -0.59 -1.43
P- value 0.3900 0.5708 0.1896
HCI -0.31 -0.25 -0.18**
standard error 0.7649 0.6191 0.0603
T- value -0.40 -0.40 -2.96
P- value 0.7004 0.7009 0.0183
B20 -0.08*** -0.07*** -0.07***
standard error 0.0118 0.0111 0.0075
T- value -6.82 -6.68 -9.83
P- value 0.0001 0.0002 0.0000
Un 0.011 0.003 0.010
standard error 0.0148 0.0169 0.0109
T- value 0.77 0.20 0.93
P- value 0.4621 0.8445 0.3794
R Square 0.93 0.92 0.92
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
158
Table 6.2(b)
OLS and Empirical Bayes Estimates for Urban Sindh
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 1.54** 1.29** 1.30***
standard error 0.5665 0.4545 0.1892
T- value 2.72 2.83 6.85
P- value 0.0264 0.0221 0.0001
Ypc -0.0001 -0.0001 -0.0001
standard error 0.0002 0.0002 0.0001
T- value -0.51 -0.28 -1.04
P- value 0.6237 0.7833 0.3286
HCI -0.39 -0.21 -0.18**
standard error 0.8363 0.6438 0.0603
T- value -0.46 -0.32 -2.97
P- value 0.6553 0.7581 0.0180
B20 -0.07*** -0.07*** -0.07***
standard error 0.0121 0.0112 0.0074
T- value -5.87 -6.05 -9.09
P- value 0.0004 0.0003 0.0000
Un 0.0008 -0.0010 0.0081
standard error 0.0327 0.0200 0.0275
T- value 0.02 -0.05 0.29
P- value 0.9807 0.9602 0.7767
R Square 0.91 0.91 0.91
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
6.2(c) Urban KPK
OLS and empirical Bayes estimate for urban KPK are given in Table 6.2(c).
It shows somewhat different results from rural KPK. Here Ypc have positive sign in case
of OLS and HP, while it is negative in case of CZ and is insignificant in all the cases.
HCI is significant and bears the expected sign in case of CZ, and is insignificant in other
cases. Unemployment shows negative relationship with BNGI in case of HP and CZ but
159
it is insignificant. The impact of B20 on the poverty level (BNGI) is significant and
according to prediction of economic theory.
6.2.(d) Urban Balochistan
Results for urban Balochistan are given in Table 6.2(d). The sign of coefficients
of per capita income, human capital index, share of income held by bottom 20%, and
unemployment are in accordance with expectations in all the three approaches adopted.
HCI is significantly related with the basic needs gap index in CZ approach. Though
unemployment bears positive sign in both rural and urban areas, but it is statistically
insignificant. The impact of B20 on the poverty level (BNGI) is strongly significant and
according to prediction of economic theory.
160
Table 6.2(c)
OLS and Empirical Bayes Estimates for Urban KPK
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 0.44 0.97* 1.18***
standard error 0.7678 0.4993 0.1881
T- value 0.5700 1.9427 6.2597
P- value 0.5843 0.0880 0.0002
Ypc 0.0002 0.00003 -0.0001
standard error 0.0002 0.0002 0.0001
T- value 0.78 0.16 -0.51
P- value 0.4601 0.8775 0.6232
HCI 0.14 -0.11 -0.18**
standard error 0.7348 0.6061 0.0600
T- value 0.19 -0.18 -3.04
P- value 0.8504 0.8604 0.0160
B20 -0.0388* -0.0559*** -0.0572***
standard error 0.0186 0.0124 0.0100
T- value -2.09 -4.50 -5.71
P- value 0.0703 0.0020 0.0004
Un 0.0029 -0.0001 -0.0002
standard error 0.0152 0.0169 0.0105
T- value 0.19 0.00 -0.02
P- value 0.8534 0.9976 0.9858
R Square 0.79 0.76 0.76
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
161
Table 6.2(d)
OLS and Empirical Bayes Estimates for URBAN Balochistan
Dependent Variable: BNGI
OLS Empirical Bayes HP
Empirical Bayes
CZ
CONSTANT 1.45** 1.24** 1.32***
standard error 0.5845 0.4643 0.1863
T- value 2.49 2.66 7.06
P- value 0.0377 0.0287 0.0001
Ypc -0.0004 -0.0001 -0.0002
standard error 0.0003 0.0002 0.0002
T- value -1.13 -0.63 -1.15
P- value 0.2909 0.5444 0.2821
HCI -0.0127 -0.0720 -0.1744**
standard error 0.6253 0.5938 0.0600
T- value -0.02 -0.12 -2.91
P- value 0.9843 0.9065 0.0197
B20 -0.0587*** -0.0638*** -0.0591***
standard error 0.0125 0.0112 0.0116
T- value -4.68 -5.69 -5.10
P- value 0.0016 0.0005 0.0009
Un 0.0191 0.0051 0.0179
standard error 0.0325 0.0200 0.0291
T- value 0.59 0.25 0.61
P- value 0.5737 0.8066 0.5558
R Square 0.84 0.83 0.83
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
162
6.3 Overall Areas
Results for overall regions are given in Tables 6.3(a) to 6.3(d). The objective of
this analysis is to investigate whether the results for decomposed sections, in rural
urban split are compatible or otherwise with the estimated results for overall areas. The
situation of BNGI in rural Pakistan is significantly different from urban areas. Some
factors that play a vital role in reducing BNGI in the urban communities have very little
effect in rural areas.
6.3 (a) Overall Punjab
The empirics for overall Punjab are given in Table 6.3(a). As far as the
relationship between per capita income and the gap between basic needs is concerned,
it is negatively related with basic needs gap index as per expectation. The t-ratio
confirms that per capita is highly significant for OLS and both for empirical Bayes (CZ)
as well as (HP). These results are more consistent with rural areas of Punjab.
HCI shows positive sign, as was in the case of rural Punjab, and is statistically
significant in OLS and CZ approaches. Due to dominance of population concentration in
rural areas, the overall results are not very different from the results of rural areas only.
Share of income held by bottom 20% follows the same pattern as it is in rural and urban
Punjab and is highly significant. An interesting finding of the study is that the rate of
unemployment is negatively related with the reduction of poverty (basic needs gap
index) in overall Punjab as was observed in the case of rural Punjab.
The behaviour of rural and overall areas of Punjab is similar. This is because the
great part of population is residing in rural areas, having a great impact on picture of
overall Punjab.
163
Table 6.3(a)
OLS and Empirical Bayes Estimates for Overall Punjab
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 1.30** 1.31** 1.61***
standard error 0.5160 0.4781 0.1353
T- value 2.52 2.74 11.91
P- value 0.03556 0.02541 0.000002
Ypc -0.0015*** -0.0007** -0.0012***
standard error 0.0004 0.0003 0.0003
T- value -3.94 -2.41 -4.43
P- value 0.0043 0.0428 0.0022
HCI 4.14*** 1.44 2.60**
standard error 1.2103 0.8709 0.9107
T- value 3.42 1.66 2.86
P- value 0.0090 0.1360 0.0212
B20 -0.07** -0.06*** -0.08***
standard error 0.0214 0.0174 0.0094
T- value -3.12 -3.58 -8.00
P- value 0.0143 0.0072 0.00004
Un -0.07* -0.03 -0.04
standard error 0.0317 0.0289 0.0260
T- value -2.16 -0.95 -1.52
P- value 0.0625 0.3692 0.1665
R Square 0.77 0.32 0.71
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
6.3 (b) Overall Sindh
The regression results for overall Sindh are given in Table 6.3(b). These results
are quite consistent in terms of sign and size with the rural Sindh. Per capita income in
both the cases is negatively related with the poverty BNGI and is highly significant. HCI
is insignificant and bears positive sign as was the case with rural Sindh. However, this is
164
contrary to the behaviour of urban Sindh. The coefficient for the variable B20 (share of
income held by bottom 20% population), is highly significant in reducing poverty.
Unemployment has negative connectivity with basic needs gap index both in rural Sindh
and overall Sindh but it is statistically insignificant.
Table 6.3(b)
OLS and Empirical Bayes Estimates for Overall Sindh
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 2.06*** 1.69*** 1.66***
standard error 0.5432 0.4692 0.1356
T- value 3.80 3.59 12.24
P- value 0.0052 0.0071 0.000002
Ypc -0.0009*** -0.0006* -0.0008***
standard error 0.0003 0.0003 0.0002
T- value -3.51 -2.25 -3.61
P- value 0.0080 0.0544 0.0069
HCI 1.36 0.68 1.38*
standard error 0.7564 0.7374 0.6299
T- value 1.80 0.92 2.19
P- value 0.1099 0.3865 0.0598
B20 -0.09*** -0.07*** -0.07***
standard error 0.0197 0.0165 0.0117
T- value -4.35 -4.47 -6.26
P- value 0.0024 0.0021 0.0002
Un -0.05 -0.03 -0.04
standard error 0.0386 0.0289 0.0360
T- value -1.31 -0.97 -1.02
P- value 0.2279 0.3591 0.3371
R Square 0.80 0.58 0.78
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
165
Table 6.3(c)
OLS and Empirical Bayes Estimates for Overall KPK
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 2.10** 1.71** 1.63***
standard error 0.8357 0.5140 0.1380
T- value 2.51 3.32 11.84
P- value 0.0361 0.0105 0.000002
Ypc -0.0003 -0.0001 -0.0001
standard error 0.0004 0.0003 0.0002
T- value -0.71 -0.51 -0.66
P- value 0.5000 0.6230 0.5294
HCI -1.04 -1.02 -0.54
standard error 0.7906 0.7344 0.5018
T- value -1.32 -1.39 -1.09
P- value 0.2229 0.2007 0.3095
B20 -0.08*** -0.07*** -0.07***
standard error 0.0235 0.0169 0.0149
T- value -3.38 -4.18 -4.73
P- value 0.0096 0.0031 0.0015
Un 0.0098 0.0119 -0.0040
standard error 0.0403 0.0292 0.0359
T- value 0.24 0.41 -0.11
P- value 0.8138 0.6937 0.9143
R Square 0.66 0.32 0.64
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
6.3(c) Overall Areas KPK
The OLS and empirical Bayes estimate for overall KPK are presented in Table
6.3(c). Per capita income in KPK has inverse relation with the basic needs gap index as
expected. This result is consistent with rural KPK but contradictory to urban KPK.
Human capital index for overall KPK is negatively related to basic needs gap index,
which is supported by theory. Again this result is consistent with rural KPK, which
166
implies that the behaviour of majority residing in rural areas dominates the urban
behaviour. The share of income held by bottom 20% is highly significant and more or
less similar to urban KPK. Unemployment rate is statistically insignificant, and the
coefficient obtained by CZ approach differs in sign when compared to OLS and
empirical Bayes (HP).
6.3(d) Overall Balochistan
The results for overall Balochistan are given in Table 6.3(d). When the three
estimation techniques are employed to data for overall Balochistan, the correlation of
two variable, namely the per capita income and share of income held by bottom 20
percent, with the dependent variable is not only strong (significant) but also according to
theory. So far as the variable human capital index is concerned, its correlation with
basic needs gap index is weak (insignificant) and also contradictory to theoretical
expectation. Once again, the results show that human capital has to do little with
poverty reduction in case of Baluchistan.
The results suggest that unemployment has a positive correlation with poverty
measured by the basic needs gap index; and this is in line with theory. The OLS gives
correct but insignificant relationship. However, in case of empirical Bayes (HP) and
(CZ), the impact is not only insignificant but even the signs are not correct.
167
Table 6.3(d)
OLS and Empirical Bayes Estimates for Overall Balochistan
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 2.10*** 1.72*** 1.67***
standard error 0.4575 0.4532 0.1338
T- value 4.59 3.80 12.48
P- value 0.0018 0.0052 0.0000
Ypc -0.0009** -0.0006* -0.0007***
standard error 0.0003 0.0003 0.0001
T- value -3.19 -2.09 -4.75
P- value 0.0127 0.0700 0.0015
HCI 0.41 0.09 0.69
standard error 0.5334 0.6938 0.4311
T- value 0.77 0.13 1.59
P- value 0.4612 0.9008 0.1505
B20 -0.069*** -0.067*** -0.066***
standard error 0.0121 0.0152 0.0115
T- value -5.72 -4.39 -5.76
P- value 0.0004 0.0023 0.0004
Un 0.0050 -0.0007 -0.0032
standard error 0.0282 0.0275 0.0263
T- value 0.18 -0.03 -0.12
P- value 0.8644 0.9802 0.9057
R Square 0.86 0.69 0.85
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
Summary of above results for the rural, urban and overall areas is given in Table
6.3(e). Where real per capita income , and share of income held by bottom 20 % has
pre-dominantly negative relationship. Human capital and Unemployment rate appeared
to be positively related with the BNGI in urban areas.
168
Table: 6.3(e) SUMMARY OF RESULTS 1
2
Explanatory Variables
Expected Theoretical Relationship with BNGI
Depicted Relationship General Relationship in Empirical Analysis
PUNJAB SINDH K P K BALOCHITAN
Rural Urban overall Rural Urban overall Rural Urban overall Rural Urban overall
Real Per capita Income
Ypc
(-)
(-) OLS (-) OLS (-) OLS (-) OLS (-) OLS (-) OLS (-) OLS (+) OLS (-) OLS (-) OLS (-) OLS (-) OLS Pre-dominant (-)
(-) HP (-) HP (-) HP (-) HP (-) HP (-) HP (-) HP (+) HP (-) HP (-) HP (-) HP (-) HP
(-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ
Human capital Index
HCI
(-)
(+)OLS (-) OLS (+) OLS (+)OLS (-) OLS (+) OLS (-) OLS (+) OLS (-) OLS (+)OLS (-)OLS (+) OLS Most Likely
(+) for urban
areas (+) HP (-) HP (+) HP (+) HP (-) HP (+) HP (-) HP (-) HP (-) HP (-) HP (-) HP (+) HP
(+) CZ (-) CZ (+) CZ (+) CZ (-) CZ (+) CZ (-) CZ (-) CZ (-) CZ (+) CZ (-) CZ (+) CZ
Share of Bottom 20% Population in
Income
B20
(-)
(-) OLS (-) OLS (-) OLS (-) OLS (-) OLS (-) OLS (-) OLS (-) OLS (-) OLS (-) OLS (-)OLS (-) OLS Pre-dominant
(-) (-) HP (-) HP (-) HP (-) HP (-) HP (-) HP (-) HP (-) HP (-) HP (-) HP (-) HP (-) HP
(-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ (-) CZ
Rate of Unemployment
Un
(+)
(-)OLS (+) OLS (-) OLS (-) OLS (+) OLS (-) OLS (+)OLS (+) OLS (+) OLS (+)OLS (+)OLS (+) OLS Most Likely (+)
for urban areas
(-) HP (+) HP (-) HP (-) HP (-) HP (-) HP (+) HP (-) HP (+) HP (+) HP (+) HP (-) HP
(-) CZ (+) CZ (-) CZ (-) CZ (+) CZ (-) CZ (-) CZ (-) CZ (-) CZ (+) CZ (+) CZ (-) CZ
169
6.4 Rural-Urban Analysis using Aggregate Prior
This section presents an empirical analysis from different angles so as to
ascertain the relationship of regressors and regressand more precisely. We analyze the
rural and urban behavior in the four provinces, however using the prior (new) obtained
from all the eight areas for Bayesian estimation only. Core objective of the present study
is to investigate and quantify the relationship of per capita income, human capital index,
share of income held by bottom 20%, and unemployment rate with the level of poverty
reflected by the basic needs gap index in different regions of Pakistan for the period
1979 to 2007-08. The results for rural-urban aggregate areas are presented in Table
6.4(a) to 6.4(h).
6.4 (a) Rural Punjab
Results for rural Punjab, as depicted in Table 6.4(a), confirm again that results
based on empirical Bayes (CZ) are unambiguously precise as indicated by smaller
values of standard errors of respective variables. Only a slight difference in values of
coefficients computed by empirical Bayes HP can be observed. Value of R-Square
indicates that fit is moderately good in all the cases.
170
Table 6.4(a) OLS and Empirical Bayes Estimates for Rural Punjab
(using aggregate rural-urban Prior)
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 1.67** 1.16 1.49***
standard error 0.5601 0.7186 0.4017
T- value 2.97 1.61 3.71
P- value 0.0178 0.1457 0.0060
Ypc -0.0011** -0.0005 -0.0008**
standard error 0.0003 0.0004 0.0003
T- value -3.32 -1.49 -3.07
P- value 0.0105 0.1747 0.0153
HCI 2.70*** 1.95* 1.88**
standard error 0.8002 0.9060 0.6589
T- value 3.38 2.15 2.86
P- value 0.0097 0.0638 0.0213
B20 -0.0755*** -0.0675** -0.0714***
standard error 0.0214 0.0230 0.0164
T- value -3.52 -2.93 -4.35
P- value 0.0078 0.0189 0.0024
Un -0.0820* -0.0389 -0.0494
standard error 0.0385 0.0504 0.0328
T- value -2.13 -0.77 -1.51
P- value 0.0658 0.4630 0.1697
R Square 0.82 0.65 0.80 ***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
The analysis of results is given at the end.
171
6.4 (b) Urban Punjab
Table 6.4(b)
OLS and Empirical Bayes Estimates for Urban Punjab
(using aggregate rural-urban Prior)
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 1.60*** 1.25* 1.50***
standard error 0.3890 0.6572 0.3181
T- value 4.10 1.90 4.71
P- value 0.0034 0.0940 0.0015
Ypc -0.0002 0.0001 -0.0002
standard error 0.0002 0.0003 0.0002
T- value -0.91 0.19 -1.07
P- value 0.3900 0.8544 0.3146
HCI -0.31 -0.34 -0.13
standard error 0.7649 0.9066 0.6298
T- value -0.40 -0.37 -0.20
P- value 0.7004 0.7210 0.8471
B20 -0.081*** -0.077*** -0.079***
standard error 0.0118 0.0187 0.0104
T- value -6.82 -4.12 -7.55
P- value 0.0001 0.0033 0.0001
Un 0.011 0.023 0.009
standard error 0.0148 0.0433 0.0137
T- value 0.77 0.54 0.67
P- value 0.4621 0.6059 0.5215
R Square 0.93 0.23 0.93 ***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
172
6.4 (c) Rural Sindh
Table 6.4(c)
OLS and Empirical Bayes Estimates for Rural Sindh (using aggregate rural-urban Prior)
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 2.11*** 1.62** 1.89***
standard error 0.4093 0.6627 0.3317
T- value 5.15 2.45 5.69
P- value 0.0009 0.0399 0.0005
Ypc -0.0011*** -0.0007* -0.0010***
standard error 0.0003 0.0004 0.0003
T- value -3.84 -1.89 -3.56
P- value 0.0049 0.0956 0.0074
HCI 1.44* 0.98 1.17*
standard error 0.7332 0.8950 0.6154
T- value 1.96 1.09 1.90
P- value 0.0852 0.3074 0.0944
B20 -0.058*** -0.059** -0.053***
standard error 0.0131 0.0192 0.0125
T- value -4.44 -3.05 -4.27
P- value 0.0022 0.0157 0.0027
Un -0.19* -0.10 -0.16*
standard error 0.0868 0.0721 0.0836
T- value -2.24 -1.40 -1.96
P- value 0.0555 0.1993 0.0860
R Square 0.82 0.70 0.81
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
173
6.4 (d) Urban Sindh
Table 6.4(d) OLS and Empirical Bayes Estimates for Urban Sindh
(using aggregate rural-urban Prior)
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 1.54** 1.15 1.41***
standard error 0.5665 0.7193 0.3954
T- value 2.72 1.60 3.57
P- value 0.0264 0.1479 0.0073
Ypc -0.0001 0.0001 -0.0001
standard error 0.0002 0.0003 0.0002
T- value -0.51 0.40 -0.69
P- value 0.6237 0.6983 0.5114
HCI -0.39 -0.33 -0.16
standard error 0.8363 0.9385 0.6515
T- value -0.46 -0.36 -0.25
P- value 0.6553 0.7307 0.8071
B20 -0.071*** -0.068*** -0.070***
standard error 0.0121 0.0187 0.0100
T- value -5.87 -3.62 -6.92
P- value 0.0004 0.0068 0.0001
Un 0.0008 0.0155 0.0034
standard error 0.0327 0.0493 0.0303
T- value 0.0249 0.3153 0.1121
P- value 0.9807 0.7606 0.9135
R Square 0.91 0.43 0.91
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
174
6.4 (e) Rural KPK
Table 6.4(e) OLS and Empirical Bayes Estimates for Rural KPK
(using aggregate rural-urban Prior)
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 2.80** 2.04** 1.66***
standard error 0.9367 0.8426 0.4798
T- value 2.99 2.42 3.46
P- value 0.0174 0.0417 0.0085
Ypc -0.0010* -0.0005 -0.0006
standard error 0.0005 0.0004 0.0003
T- value -2.10 -1.14 -1.72
P- value 0.0688 0.2891 0.1244
HCI -1.44 -1.23 -0.60
standard error 0.7864 0.9017 0.5591
T- value -1.83 -1.37 -1.07
P- value 0.1043 0.2087 0.3149
B20 -0.060* -0.053* -0.029
standard error 0.0323 0.0265 0.0230
T- value -1.87 -1.99 -1.27
P- value 0.0983 0.0814 0.2413
Un 0.0205 0.0208 -0.0166
standard error 0.0517 0.0552 0.0455
T- value 0.40 0.38 -0.36
P- value 0.7017 0.7163 0.7252
R Square 0.51 0.39 0.42
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
175
6.4 (f) Urban KPK
Table 6.4(f) OLS and Empirical Bayes Estimates for Urban KPK
(using aggregate rural-urban Prior)
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 0.44 0.27 0.97*
standard error 0.7678 0.8204 0.4443
T- value 0.57 0.33 2.18
P- value 0.5843 0.7476 0.0605
Ypc 0.0002 0.0004 0.00003
standard error 0.0002 0.0003 0.0002
T- value 0.78 1.24 0.17
P- value 0.4601 0.2494 0.8699
HCI 0.14 -0.05 -0.20
standard error 0.7348 0.8906 0.5127
T- value 0.19 -0.05 -0.39
P- value 0.8504 0.9580 0.7048
B20 -0.0388* -0.0402* -0.0498***
standard error 0.0186 0.0217 0.0147
T- value -2.09 -1.86 -3.38
P- value 0.0703 0.1007 0.0096
Un 0.0029 0.0164 0.0034
standard error 0.0152 0.0435 0.0144
T- value 0.19 0.38 0.23
P- value 0.8534 0.7159 0.8209
R Square 0.79 0.06 0.77
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
176
6.4 (g) Rural Baluchistan
Table 6.4(g) OLS and Empirical Bayes Estimates for Rural Balochistan
(using aggregate rural-urban Prior)
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 1.93*** 1.52** 1.68***
standard error 0.4960 0.6957 0.3745
T- value 3.90 2.19 4.48
P- value 0.0046 0.0604 0.0021
Ypc -0.0010** -0.0006 -0.0008***
standard error 0.0003 0.0004 0.0002
T- value -3.22 -1.66 -3.41
P- value 0.0123 0.1350 0.0092
HCI 0.019 0.054 0.118
standard error 0.6928 0.8722 0.5902
T- value 0.0273 0.0622 0.2000
P- value 0.9789 0.9519 0.8465
B20 -0.047** -0.047** -0.043**
standard error 0.0153 0.0200 0.0136
T- value -3.07 -2.34 -3.18
P- value 0.0154 0.0475 0.0130
Un 0.032 0.038 0.033
standard error 0.0353 0.0502 0.0331
T- value 0.92 0.75 0.98
P- value 0.3841 0.4726 0.3542
R Square 0.72 0.64 0.72
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
177
6.4 (h) Urban Baluchistan
Table 6.4(h) OLS and Empirical Bayes Estimates for Urban Balochistan
(using aggregate rural-urban Prior)
Dependent Variable: BNGI
OLS Empirical Bayes HP Empirical Bayes CZ
CONSTANT 1.45** 1.11 1.39***
standard error 0.58 0.73 0.40
T- value 2.49 1.51 3.46
P- value 0.0377 0.1684 0.0086
Ypc -0.0004 -0.0001 -0.0004
standard error 0.0003 0.0004 0.0003
T- value -1.13 -0.24 -1.25
P- value 0.2909 0.8149 0.2472
HCI -0.013 -0.090 0.054
standard error 0.63 0.85 0.50
T- value -0.02 -0.11 0.11
P- value 0.98 0.92 0.92
B20 -0.059*** -0.056** -0.058***
standard error 0.0125 0.0189 0.0121
T- value -4.68 -2.96 -4.80
P- value 0.0016 0.0181 0.0013
Un 0.019 0.031 0.017
standard error 0.0325 0.0489 0.0306
T- value 0.59 0.63 0.56
P- value 0.5737 0.5478 0.5893
R Square 0.84 0.64 0.84
***Significant at 1% level. ** Significant at 5% level.* Significant at 10% level.
178
In the tables [6.4.(a) – 6.4(h)], results for aggregate rural-urban areas are given with
more or less the same results; however few points are worth mentioning:
OLS results for different regions as expected remain same.
Results obtained by empirical Bayes (HP) and empirical Bayes (CZ) slightly
differ, but are not statistically significant.
Results for rural areas are mostly the same in size and level of significance even
in case of both HP and CZ techniques (except rural Balochistan).
In case of urban areas results differ substantially in magnitude and even sometimes
coefficients bear inverse signs. However such coefficients in both categories are
insignificant.
6.5 Sensitivity Analysis
The requirement of a good research is to measure both dependent and
independent variables without any errors. This implies that statistic on the variables of
model used are accurate. It is assumed that they are not hypothetical estimates,
interpolated, extrapolated or smoothed by any systematic method. In practical exercise
such a situation is rare. We may face non response errors, reporting errors and
computing errors. Error in measurement gives rise to specification bias. If error of
measurement is in dependent variable, OLS estimates are still unbiased. However
variances and standard errors are different. This means that variances and standard
errors with error in measurement are larger than in the case where there are no such
errors in measurements. In case, errors measurement is in the regressors, OLS
179
estimates are not only biased but also they are inconsistent, that is, they remain biased
even if the sample size increases indefinitely.
A good deal of research is that it does not contend with results and findings
obtained by using one estimation technique. It also does not hinge upon one modeling
aspect of research that it has compact ways to compare the results. According to
Hendry and Richard (1983), the results of a research can be termed satisfactory if
research is based on following criteria.
Results support the theory, Data is coherent, Model used by a particular study
encompasses all rival models, and Regressors are uncorrelated with stochastic
term.
However, the researcher is liable to commit various specification errors. One of
such errors is the measurement errors. For the case of exposition and computational
expediency, it is assumed that dependent as well as independent variables are
measured without any error. In practice a researcher encounters various errors in
measurement because of the simple fact that computing errors as well as reporting
errors creep in due to manipulation of data and there are grave consequences of errors
of measurements.
Keeping in view these concerns of errors of measurement the present study
analyses and compares the initial results with the results generated from data which
underwent some sort of manipulation.
Human capital index (HCI) is obtained by simply taking average of educational
attainment index (EAI) and health status indices (HSI):
180
HCI2
EAI HSI
OR
HCI 0.5* 0.5*EAI HSI
Since the weights are arbitrary, someone might think of assigning more weight to EAI
than HIS or vice versa. Our target is to analyze impact of such procedure in our results.
Therefore we change the weights of EAI and HSI, and obtain HCI-1 and HCI-2 as given
below,
HCI-1 0.45* 0.55*EAI HSI
HCI-2 0.55* 0.45*EAI HSI
Using the above mentioned HCI, HCI-1 and HCI-2, we find estimates by using
three different techniques, i.e. OLS, Empirical Bayes (HP), and Empirical Bayes (CZ).
These estimates are given in Appendix I-A to Appendix IV-H.
Appendix I-A, shows the comparison of results for OLS and empirical Bayes estimates
for rural Punjab. Results for constant term are more or less same. No changes are
palpable with the exception that empirical Bayes (HP) turns significant in both HCI-1
and HCI-2 cases. Value and sign of the coefficient of Ypc also remain about same in all
three cases, however in case of HCI-1, significance level increases in all three
techniques.
Coefficient of HCI is nearly unchanged. This variable is significant at 99% for empirical
base (CZ) in case of HCI-1 and at 95% in case of initial and HCI-2. B20 is highly
significant in both HCI-1 and HCI-2. This shows that a change in HCI to the extent of
181
about five percent both upward and downward, does not lead to change the initial result
drastically. It is indicative from the value of R-squared that regression fits equally good
in all three cases. From results reported in Appendix I-A, it can safely be concluded that
if the data on HCI is manipulated upward or downward, results do not tend to change all
along. If there had been any errors of measurement in one of the explanatory variable,
that is HCI, the results would have undergone a substantial change. So it can be
asserted with certainty that data for this segment of research is reasonably coherent
and all regressors particularly HCI are weak, implying that regressors are uncorrelated
with stochastic disturbance term.
Appendix I-B also shows somewhat same picture where magnitude and sign of
coefficient are almost same except slight difference of significance in only one case i.e.
HCI-1 Empirical Bayes (CZ). In Appendix I-A, most of the time, value of the coefficient,
sign and significance also remain same. Only B20 gets significant in Empirical Bayes
(HP) in both cases. Appendix I-A shows results for the rural Balochistan, where slight
difference in significance of the variable is observed. In case of HCI-1 and in Empirical
Bayes (HP), human capital changes the sign but this variable is insignificant in all three
cases. Value of the R squared is almost same in all situations.
Appendix II-A to Appendix II-D contain results for Urban areas and here also results
show same pattern and most of the time sign and magnitude remain the same while
slight difference in significance is observed in some cases. Human capital also has the
same picture and in urban Balochistan it varies sign. In both cases and with three
econometric techniques coefficients of HCI remain significant as it was in initial results
with OLS and Empirical Bayes (H.P).
182
In cases of overall provinces, results are quoted in Appendix III-A to Appendix III-
D. Coefficients of all the variables have same sign and almost same magnitude. Only
slight difference in significance level is observed. The value of R-Squared is almost
same in most of the cases.
Results for the rural and urban areas are given in Appendix IV-A to Appendix IV-
H. In this case we observed the same pattern as it was in the case of separate rural and
urban case.
From the results presented above, it is clear that the estimates of the techniques
applied in this study are robust and do not change significantly. If the weights of EAI and
HSI are changed in the composition of HCI, the magnitude and direction of the
coefficients of various techniques remain same. However, little difference was observed
in the significance level of the variables in the study. Therefore we conclude that
redefining HCI with different weights of EAI and HSI will not change the results
presented in the study.
183
CHAPTER 7
SUMMARY & CONCLUSIONS
This chapter contains three sections. In the first section, a brief overview of the
whole exercise is presented. The second section precisely explains the main findings of
the study regarding the basic needs fulfillment situation prevalent in different parts of
Pakistan. It also covers the rural and urban comparison and factors responsible for the
gaps in basic needs fulfillment. Section-3 presents the conclusions and some policy
recommendations briefly.
7.1 Overview of the Study
The present study aimed at discussing the operational implications of basic
needs and exploring the possible and significant correlates of basic needs in Pakistan.
It contributed towards the computation of basic needs gap indices (BNGI) for the first
time in Pakistan and analyzed various factors that could possibly affect the depth and
breadth of poverty as reflected by the BNGI. The research covers four provinces of
Pakistan with rural and urban bifurcation (thus total eight regions) over the period 1979
to 2008. Other areas like Azad Kashmir, FATA, and Gilgit-Baltistan are not included in
the study due to non availability of consistent data for these areas. The previous studies
based on cross country empirical analysis faced the main criticism of having no
common yardstick to measure basic needs for all the countries due to differences in the
socio-economic and cultural situations. However, this study is free from this sort of
observation and the basic needs yardstick is presumably same for all areas of Pakistan
due to similar socio-economic, cultural and political conditions. Initially, ten variables
184
were considered that could be assumed to affect BNGI one way or the other. However,
after applying different techniques and keeping in view the theoretical considerations,
only four variables appeared to be more relevant for detailed empirical analysis. These
were the per capita income of the households, human capital index, share of income
held by the bottom 20 % and unemployment rate.
The study considered food, clothing, shelter, health, and education as the basic
needs and derived the necessary information from the official sources like the Economic
Survey of Pakistan, the HIES and PSLM, the Labour Force Survey and some other
publications. Some composite variable like the human capital index, quintiles of income
distribution and the BNGI were constructed from a variety of other variables. The data
used in this analysis covered a period of 1979-80 to 2007-08, comprising 13
observations over time and 4 observations over the cross section. The HIES information
is available generally after 3 years.
After disappointment from the results of (neoclassical) growth models and the
emergence of endogenous growth theory, human capital received tremendous attention
in the research work along with physical capital. Proponents of the endogenous growth
theory held the view that well trained and educated persons were more productive and
capable of using the new technology efficiently. Consequently, the returns to investment
(in human and physical capital) will be increasing rather than decreasing. This
investment in human capital will lead to higher productivity of labour, which in turn will
result into substantial increase in income and remunerations. Therefore, we
incorporated human capital as an important determinant of BNGI along with other usual
185
variable like the per capita incomes, the unemployment rate and the share of bottom
20% population in the distribution of income.
After descriptive comparison of BNGI among different regions and particularly
between rural and urban areas, the model was empirically analyzed through appropriate
techniques, following different estimation techniques given below:
Ordinary least square (OLS).
The Empirical Bayesian, with two alternative approaches:
Approach due to Hsiao and Pesaran (2004)
Approach due to Carrington and Zaman (1996)
7.2 Summary of Findings
Average monthly income per household is one of the determinants of basic
needs gap index. It has generally increased over the years 1984-85,1985-86, 1986-87,
1990-91, 1998-99, 2004-05 and 2005-06, but showed downward trend for the remaining
years.
Income gap between the rural and urban areas is visible, and remained a
permanent feature over the whole study period. Urban Sindh and urban Balochistan
showed high growth in income over the period concerned.
The percentage distribution of monthly income among households by quintiles for
the different regions of Pakistan also shows a dismal situation. The general picture
shows a downward trend for the share of the first quintile (lowest 20% of population);
186
however this decrease is very large and consistent in the urban areas. In contrast, an
increase in the share of income for this quintile is observed for the rural areas of Sindh
and Balochistan. In case of the highest quintile (top 20 % population), an increase in
income share has generally been observed overtime for most of the regions. These
facts about income disparities are also observable when one looks at the relative share
in incomes of the top 20% population (richest) to the bottom 20% (poorest).
Construction and analysis of the performance of basic needs indices for different
regions of Pakistan is also integral part of this study. It is obvious from the respective
Tables and Figures given in chapter-4, that initially all the regions started from more or
less the same level of basic needs. After that, till 1986-87, most of the regions showed
downward trend or at least remained below the urban regions. In the subsequent period
till 1998-99, there are fluctuations in basic needs gap index in both rural and urban
regions and generally the gap increased during this period.
Beyond 1998-99, there is a clear and visible split in rural and urban areas and
this trend persists till the end. This factual observation that inequality in urban Pakistan
remained high as compared to the rural areas has also been made by other studies, for
instance, Idrees (2006). During this spell, rural areas observed a decline in basic needs
gap index while urban areas witnessed converse trend, obviously due to constant
migration of people to big cities and the prevalence of open unemployment in the urban
areas. This gap widened till the end the study period except in rural KPK, where both
rural and urban areas show an increase in BNGI after 2001-02. Situation in urban Sindh
187
and urban KPK is alarming where a persistent increase in BNGI is observable since
1993-94.
The health index and education attainment indices lead to the construction of the
human capital index (HCI). The mean human capital for rural areas turns out to be 0.46
and that for urban areas to be 0.64, obviously due to better health and education
facilities available in the urban areas. There is continuous rise in HCI for almost all the
regions; however, this rise is significant during the period 1985-86 to 1993-94. The
provinces can be ranked in the following sequence with reference to human capital
index: Sindh (highest), Punjab, KPK, and Balochistan (lowest). At provincial level,
however, this difference is not persistent. For instance, Punjab and KPK followed the
same trend as that of Pakistan. However, significant progress in human capital is visible
in rural KPK, where in 1979 value of HCI was 0.33 and in 2007-08 it appeared as 0.58.
The main objective of the present study is to investigate and quantify the
correlation of various variables with BNGI in different regions of Pakistan. After
exclusion of certain less significant and impotent variables, the final model includes four
variables as discussed above, namely, the per capita income, human capital index,
share of income held by bottom 20%, and unemployment rate.
Empirical Bayes (CZ) gives precise results when compared to OLS and empirical
Bayes (HP). This fact can be observed in Tables (6.1a - 6.4h), where standard error is
smaller in case of CZ as compared to OLS and HP.
The result reveal that per capita income and BNGI are negatively correlated in
both rural and urban areas of Pakistan. Although this is in accordance with our prior
188
expectations; however for most of the time this impact is insignificant for urban areas
but significant for rural areas.
The results obtained for aggregate rural and urban area slightly differ, especially
for those variables, which are statistically not significant. Results for rural areas are
mostly same in size and level of significance even in case of both HP and CZ
techniques (except for rural Balochistan). However, in case of urban areas, the results
differ substantially in magnitude and even sometimes the coefficients bear inverse
signs. Therefore, our main focus remained on rural, urban and overall regions.
HCI is negatively related to BNGI in urban areas and supports the prior
information. This variable is statistically significant for empirical Bayes (CZ) in all urban
areas. This implies that due to an improvement in HCI, people get more opportunities to
earn and fulfill their basic needs. HCI shows mixed relationship with BNGI in rural areas
except KPK, where there is a direct relationship between human capital index and
BNGI. In rural Punjab and rural Sindh, it is statistically significant.
The share of income held by bottom 20% appeared as one of the strong factors
that is negatively correlated with the BNGI. This shows that an increase in share of
income held by bottom 20% and the betterment of the masses in terms of basic needs
fulfillment, go hand in hand. For most of the events, this variable is highly significant for
both the rural and urban areas.
The rate of unemployment shows mixed results. In case of urban areas, its
relationship is positive with BNGI, which is in accordance with the prior expectations;
189
while in rural areas, except rural Balochistan, it negatively affects BNGI for obvious
reasons. However, this variable appeared statistically insignificant in most of the cases.
7.3 Conclusions and Policy Recommendations
After the failure of growth oriented approach in the late 1970s, there evolved a
consensus among the researchers that benefits of growth should be passed on to the
poor. Redistribution of income and accelerated economic growth is now considered as
policy prescription for the developing countries to eradicate poverty and hunger from the
world. At the international level, the Millennium Development Goals (MDGs) have been
formulated by the United Nations. According to this declaration, everyone is sovereign,
has the right to live with dignity, and the rights to meet the basic standards of living.
First goal of this declaration is to eradicate extreme poverty and hunger. Two
main targets to achieve this goal are:
1) Reduce poverty between 1990 and 2015 by 50% (in 1990 proportion of population
below the poverty line was 26.1% and target for 2015, is 13%).
2) Reduce the proportion of people who suffer from hunger by 50% by this period.
Poverty reduction received great attention as the key objective of development
policies. However, due to its multidimensional nature, no final definition and
measurement is agreed upon. Despite this serious shortcoming, the present study
employed a popular method to measure poverty via the yardstick of basic needs gap
indicator for different regions of Pakistan. This research is unique of its nature in this
respect.
190
The study found tangible gaps in income, expenditure and savings, among rural
and urban areas of Pakistan. Gap between the highest and the first quintile is widening
in most of the regions, which is not good for the social fabric of the country. In particular,
this uneven distribution is mostly observed in the urban areas.
All regions of Pakistan witnessed a substantial and consistent increase in per
capita income during study period. Surprisingly this growth in GNP per capita led to
worsening of distribution. This phenomenon supports the Kuznets hypothesis that
distribution of income worsens at early stages of economic development. Inequality
persisted and remained unabated in Pakistan due to the fact that our economy has not
yet attained the climax after which inequality will tend to decline. A prompt political
dispensation can curtail the acute problem of poverty even in the presence of growing
inequalities. Such political will is needed from policy makers, government official and
political stalwart of the country.
Based on its findings and analysis, the study will put forward the following
plausible and concrete suggestions and policy implications:
1. Human capital appears to be a strong and significant factor to make dent in
poverty on one side and to act as a natural catalyst for growth and equity in distribution
of income. The central policy must emphasize on creating demand for innovation and
indigenous capability building technology. This requires Pakistan to allocate appropriate
allocation for higher education. Access to quality education is instrumental to strengthen
scientific infrastructure. Special attention needs to be given to reduce large urban-rural
and gender disparities in enrollment rates at the secondary level. Strategies are also
191
needed to improve the educational profile of poor so that they can take advantage of
increasing opportunities in the job market available across the globe.
The government needs to focus on providing good quality basic health services
particularly in rural areas of Pakistan in which the condition of masses is miserable. The
present bias against poor regions and between rural and urban areas needs to be
reversed. Government is required to enter into operational partnerships with local
communities and NGOs to provide health care to poor segment of society.
Unemployment rate is an important factor responsible for poverty and a hurdle
towards fulfillment of basic needs. Government needs to revise its strategies on
patterns that encourage labor intensive growth. Such activities in agriculture,
manufacturing and export oriented sectors need to be encouraged so as to generate
more employment and ensure the spillover effects of growth for poor. GDP growth per
employed person and employment rate are among targets to achieve the first objective
(i.e. to eradicate extreme poverty and hunger) of MDGs. The positive relation between
unemployment and BNGI is observed in rural areas of Pakistan. Besides data issues,
disguised employment in villages is one cause of this positive relationship. Hence it is
recommended to create job opportunities in the rural areas so that masses could meet
their basic needs and to discourage further migration to big cities, which is now creating
grave socio-economic and political problems.
Economic growth remained respectably appropriate during the first six-seven
years of last decade. However, this robust economic growth remained concentrated in
relatively skilled-intensive sectors of finance, telecommunication and information
192
technology (IT), which favored only those with high income who could afford higher
education. Unskilled workers and those with low education level failed to integrate
themselves in the sectors of high growth during this period. Sectors like housing and
construction, SMEs are relatively labor intensive with high employment elasticity. These
sectors need to be given special priority for rapid expansion to ensure some gain to the
poorer sections of the society. Direct attack on poverty and inequality needs to reinforce
the social security nets that are already in progress.
On one hand, increased health facilities and innovations has enhanced the life
expectancy considerably; but on other hand the increased food production due to
technological advancement has failed to address the problems of hunger and
malnutrition in developing countries. The study has found a wide and persistent gap
between rural and urban segments of the society in terms of human capital. This bias
not only creates social and moral degradation and so many other evils in the society,
but also causes further disparity in the region. Unplanned urbanization makes the cities
overcrowded, which lack adequate infrastructure to accommodate influx of labour to
urban regions. When observed in terms of interregional disparity, the study found that
Balochistan is lagging behind other provinces in human capital index and this trend is
prevalent in both rural and urban areas.
Different regions of Pakistan have some differences on socio-cultural and
political basis. These differences are natural, but prolonged differences in economic
conditions among and within regions are posing some serious problems. Pakistan has
already paid the price in 1971. The sense of economic deprivation and exclusion is also
193
dangerous that is liable to bring about so many social evils in the society. Growth for the
sake of growth is meaningless unless it reduces the suffering and miseries of the
masses. To make every person part of development process, it needs to ensure that no
one is deprived and marginalized.
Research in economics greatly relies on data but reliable and disaggregated data
for spatial analysis in most of the developing countries is not available. Although this
study covered provinces of Pakistan with rural urban bifurcation, however there is still
scope to investigate these issues at district level. We can hope for availability of quality
data that is more dis-aggregated in future, which will enhance the quality of research.
194
REFERENCES
Adams, R. H. and Page, J. (2005). Do international migration and remittances reduce
poverty in developing countries? World Development, 2005, Vol. 33, no. 10, p.
1645-1669.
Adelman, I, & C. T. Morris. (1971). Economic growth and social equity in developing
countries. Stanford, CA: Stanford University Press.
Afxentiou, P.C. (1990). The rhetoric and substance of basic needs. South African
Journal of Economics, 58(1), 45-50.
Ahluwalia, M.S. (1974). Income equality some dimensions of the problem. In H chenery
et al ; Redistribution with growth. Oxford : Oxford university press.
Ahluwalia, M.S, G. C. Nicholas, and Chenery, H.B (1979),Growth and Poverty in
Developing Countries. Journal of Development Economics. No. 16.
Ahmad, Ehtisham. And Stephen. Ludlow. (1989). Poverty Inequality and Growth in
Pakistan, The Pakistan Development Review, 28:4, 831-850.
Alauddin, T. (1975), Mass Poverty in Pakistan: A further Study. The Pakistan
Development Review, Volume 14(4), pp. 431-450.
Alesina, Alberto and Dani Rodrik (1993), income distribution and economic Growth. A
simple theory and some empirical evidence, The political economy of Business
Cycles and growth (MIT Press, Cambridge, MA).
195
Amjad, R., & Kemal, A .R. (1997). Macroeconomic policies and their impact on poverty
alleviation in Pakistan. The Pakistan Development Review, 36(1), 39-68.
Anand, Sudhir and S.M. Ravi Kanbur, (1984), The Kuznets Process and the Inequality-
Development Relationship, Journal of Development Economics No. 23.
Anderson, E., & White, H. (2001). Growth versus distribution: Does the pattern of
growth matter? Development Policy Review, 19(3), 267–289.
Anoruo E., and Ahmad, Y. ( 2001), Causal Relationship between Domestic Savings and
Economic
Anwar, Talat. (2003). Trends in Inequality in Pakistan between 1998-99 and 2001-02,
The Pakistan Development Review 42:4 Part II. pp 809-821.
Arif, G. M. and Bilquees, Faiz. (2007). chronic and transitory poverty in Pakistan:
Evidence from a longitudinal Household Survey, The Pakistan Development
Review. 46;2 pp 111-127.
Arvind, V. (2006). Poverty and hunger in India: What is needed to eliminate them.
Pakistan Development Review, 45(2), 241-259.
Ayub, M.A. (1977). Income Inequality in a growth theoretic context. The Case of
Pakistan. PhD Thesis Submitted to Yale University.
Bacha, E.L., 1990, A Three-Gap Model of Foreign Transfers and the GDP Growth Rate
in Developing Countries, Journal of Development Economics, Vol. 32, 279-96.
196
Banerjee, A. V., & Duflo, E. (2006). The economic lives of the poor. Bread Working
Paper, 135.
Banerjee, A.V. and Newman, A. F.(1993), Occupational choice and the Process of
Development – Journal of Political Economy No. 101.
Galz. O. and Zaira, J. (1993), Income Distribution and Macroeconomics, Review of
economic Studies No. 60.
Bergan, A. (1967): Personal Income Distribution and Personal Savings in Pakistan,
1963-64.The Pakistan Development Review, Vol 7:160-212.
Berger, J. O. (1985). Statistical decision theory and Bayesian analysis. New York:
Springer-Verlag.
Blattberg, R. & George, E. I. (1991). Shrinkage Estimation of Price and Promotional
Elasticities: Seemingly Unrelated Equations, Journal of the American Statistical
Association, 86, 304-315.
Blaug, M. (1972). The correlation between education and earnings: What does it
signify? Higher Education,1(1).
Campos,R and Palomo. Anabella Lardé de (2002). “Invirtamos en Educación para
Desafiar el Crecimiento Económico y la Pobreza,” Fundación Salvadoreña para
el Desarrollo Económico y Social, San Salvador (May).
197
Caroll, C. D, Overland. J, Weil. D. N. (2000). Saving and Growth with Habit Formation,
American Economic Review, Vol. 90, 3:351-55.
Carrington, W. J. & Zaman, A. (1994). Interindustry variation in the costs of job
displacement. Journal of Labor Economics, 12(2), 243-275.
Cheema, I. A. (2005). A profile of poverty in Pakistan: Senior poverty specialist, centre
for research on poverty reduction and income distribution. Islamabad: Planning
Commission Pakistan.
Chimhown, A. O, Piesse. J, Pinder. C. (2005). The Socio-economic Impact of
Remittances on Poverty Reduction. In Maimbo SM, D Ratha (eds.). Remittances:
Development, Impact and Future Prospects. Washington D.C: The World Bank.
Dagdeviren, Hulya, Rolph van der Hoven, and John Weeks (2002). Poverty with Growth
and Redistribution,’ Development and Change 33, 3 pp 383-413.
Dasgupta. P. and Ray. D(1987). Inequality as a determinant of malnutrition and
unemployment policy. Economic Journal No. 97.
Datt, Guarav.(2002). Implementation completion report, Philippines: enhanced poverty
monitoring- studies component. Washington, DC: World Bank.
Deaton, A. (2004). Measuring poverty: Research program in development studies.
Working paper. New Jersy: Princeton University
198
Defina, R. (2001). The Impact of Macroeconomic Performance on Alternative Poverty
measures, Social Science Research. 31: pp 29-48.
DeGregorio, J. (1992). Economic Growth in Latin America, Journal of Development
Economics, Vol. 39: 59-84.
Deininger, K., & Squire, L. (1996). Measuring income inequality: A new data set. Oxford
Journal of Social Sciences, World Bank Economic Review, 10 (3), 565-591.
Dreze, Jean, and Amartya. K. Sen. (1989). Hunger and Public Action. Oxford:
Clarendon Press.
Edwards, S. (1995). Why are Saving Rates so Different Across Countries?: An
International Comparative Analysis, NBER Working Papers No.5097.
Elahi Mahboob, Some Poverty issues in Pakistan, an empirical analysis. Unpublished
paper.
Emmanuel Anoruo1, Yusuf Ahmad (2001).Causal Relationship between Domestic
Savings and Economic Growth: Evidence from Seven African Countries. African
Development Review. Volume 13, Issue 2, pages 238–249.
Federal Bureau of Statistics. (1979 to 2008). Household integrated economic survey.
Islamabad: Author.
Field, G. S. (1989). changes in poverty and inequality in the developing countries
(Mimeographed).
199
Fisher, I. (1906). The Nature of capital and income. Reprints of Economic Classics,
(Reprinted 1965). New York: Augustus M. Kelley.
Geweke, J. (2005). Contemporary Bayesian econometrics and statistic. (Wiley series
in probability and statistics). New York: Wiley.
Ghai, D. P., Khan, A. R , Lee, E . L .H & Alfthan, T. (1980). Basic needs approach to
development: Some issues regarding concepts and methodology. International
Labor Organization. Switzerland, Lightning source Inc. 124.
Green, W. H. (2003). Econometrics analysis, 5th ed. New Jersey: Prentice- Hall.
Gujrati, N. Damador. (2003). Basic Econometrics 4th ed. Mc Graw Hill/ Irwin New York
pp-516
Haider, S. S. (2006). Economic analysis of crime, punishment and deterrence: An
empirical investigation for Pakistan. Unpublished PhD thesis, Quaid-i-Azam
University, Islamabad.
Haq, Rashida. (1998). Trend in Inequality and Welfare in Consumption Expenditure:
The Case of Pakistan. The Pakistan Development Review, Vol 37(4), pp 623-
664.
Haq, K. (1964). A Measurement of Inequality in Urban Personal Income Distribution in
Pakistan. The Pakistan Development Review, 4:4, 623-664.
200
Hasan, Z. (1997). Fulfillment of basic needs: Concept, measurement, and Muslim
countries performance. Journal of Economics and Management 2 (5), 1-
38.
Hicks, N. L., & Streeten. P. (1979). Indicators of development: The search for a basic
needs yardstick. World Development, 7 (6), 567-580.
Hicks, N, L.(1979), Growth versus basic needs: is there a trade off. World Development
Report.
Hopkins, M., & Hoeven, R.V.D. (1983). Basic needs in development planning. England:
Gower.
Hsiao, C., & Pesaran, M.H. (2004). Random coefficient panel data models. Institute for
the Study of Labor.(IZA), Discussion Paper, Bonn.
Idrees, Muhammad. (2006). An Analysis of Income and Consumption Inequalities in
Pakistan. Unpublished PhD thesis. Department of Economics Quaid-i-Azam
University, Islamabad.
Imran Sharif Chaudhry, Shahnawaz Malik, and Abo ul Hassan. (2009). The Impact of
Socioeconomic and Demographic Variables on Poverty: A Village Study. The
Lahore Journal of Economics, 14 : 1, pp. 39-68.
Irfan, M. (2007). Poverty and natural resource management in Pakistan. The Pakistan
Development Review. 46 (4), 691-708.
201
Jamal, H. (2003) Poverty and Inequality during the Adjustment Decade : Empirical
Findings from Household Surveys. The Pakistan Development Review. 42:2, pp
125-135.
Jamal, H. (2006). Does inequality matter for poverty reduction? Evidence from
Pakistan’s poverty trends. The Pakistan Development Review. 45 (3), 439–459.
Japelli, T, and Pagano. M. (1994). Savings, Growth and Liquidity Constraints, Quarterly
Journal of Economics, 109: 83-109.
Jeffri, S, M. Younas, and A. Khattak. (1995). Income Inequality and Poverty in Pakistan.
The Pakistan Economic and social Review, 33:1, 165-193.
Jehle, A. Geoffrey. (1990). Inequality in Pakistan: A Sectoral Welfare Approach.
Pakistan Journal of Applied Economics. 2: 165-193.
Kaldor, N. (1956). Alternative Theories of Distribution, Review of Economic Studies,
23(2): 83.
Khan. S. Rafi. (1999). Fifty Years of Pakistan’s Economy, Oxford University Press
,,Karachi,Pakistan
Khandkar, R. (1973). Distribution of Income and wealth in Pakistan. Pakistan Economc
and Social Review.
202
Korea Institute for Health and Social Affairs (KIHASA) (2000). Low Fertility and Policy
Responses to Issues of Ageing and Welfare. Research Paper 2000-1. Seoul:
Korea Institute for Health and Social Affairs and United Nations Population Fund.
Kipanga, A. A. (2007). Basic needs gap in developing economies: measurement and
determinants. Unpublished PhD thesis submitted to International Islamic
university Malaysia.
Kravis, I. B. (1960). International differences in distribution of income. Review of
Economics and statistics, 42, pp 408-416.
Kruijk, Hans de (1986) nequality in the Four Provinces of Pakistan. The Pakistan
Development Review, 25:4, pp 685-706.
Kuznets, S. (1955). Economic growth and income inequality. American Economic
Review, 45(1), 1-28.
Lanjouw, Peter, and Ravallion. Martin. (1995). Poverty and Household size, The
Economic Journal,105 (November),pp 1415-1434.
Lewis, W. A. (1954). Economic development with unlimited supplies of labour.
Manchester School of Economic and Social Studies. 22(2), 139–191.
Lewis, W. A. (1955). The Theory of Economic Growth. Homewood, III: Irwin.
Mahmood. Zafar. (1984). Income Inequality in Pakistan; An Analysis of Existing
Evidence. The Pakistan Development Review, 23; 2&3, pp 365- 376.
203
Marshall, A. (1890). The principles of political economy, a 1994 reprint of the 8th ed.,
from 1920. Hong Kong: Macmillan Press Ltd.
Meir, Gerald M and Rauch. James. E. (2002). Leading issues in Economic
Development . 7th Edition Oxford University press, Inc New York. pp-18.
Milanovic, & Yitzhaki. (2002). Decomposing world income distribution: Does the world
have a middle class ? Review of Income and wealth, 48 (2), 155-178.
Mohan, Ramesh, (2006), Causal relationship between savings and economic growth in
Countries with different income levels. Economics Bulletin, Vol. 5, No. 3, pp. 1-12
Montek, Ahluwalia. (1974). Income inequality, some dimensions of the problem. Oxford
university press land.
Naseem, S.M. (1973). Mass Poverty in pakistan: Some Preliminary Findings. The
Pakistan Development Review, Vol 12:4, 312-360.
Navaratnam, R. (2003). Malaysia's economic challenges: A critical analysis of the
Malaysian economy governance and society. London: ASEAN Academic Press.
Njimanted, Godfrey Forgha (August 2006), Econometric Model of Poverty in Cameroon:
A System Estimation Approach. International Review of Business Research
Papers. Vol.2. No. 2., pp. 30-46
204
Nurudeen, ABU. (2010). Saving-Economic Growth nexus in Nigeria, 1970-2007:
Granger causality and co-Integration analyses. Review of Economic and
Business studies. Volume 3, Issue 1, pp. 93-104.
Oshima, H. (1962). International comparison of size distribution of family incomes with
special reference to Asia. Review of Economics and Statistics, Vol.44: 439-45
Pankert, F. (1973). Income Distribution at different levels of Development: A Survey of
evidence, International Labour Review No. 108.
Papanek, G. and Kyn, O. (1986), The effect on income distribution of development, the
growth rate and economic strategy. Journal of Development Economics No. 23.
Papanek, G.(1987),Flattening the Kuznets Curve: The Consequences for Income
Distribution of Development Strategy, government intervention, income and the
rate of growth – The Pakistan Development Review No 26 (1987), Also World
Bank Development Report 1990.
Park, A. Wang. S, and Wu. G. (2002). Regional Poverty Targeting in China. Journal of
Public Economics, Vol 86 pp 123-153.
Parotti, R.(1996), Growth, Income Distribution and Democracy: What the datasay.
Journal of Economic Growth.
205
Pasha, H.A, and Altaf. M.A. (1987). Return Migration in a Life-style Setting: An
Exploratory Study of Pakistani Migrants in Saudi Arabia. Pakistan Journal of
Applied economics, 6(1), pp 1-21.
Planning commission of Pakistan (Various Issues), Economic Survey of Pakistan,
Government of Pakistan. Printing corporation of Pakistan, Islamabad.
Poverty Briefing No.7 (1999), Unemployment and Poverty.
Prottori, Roberto. (1996). Income Distribution Democracy and growth. What the data
say, Journal of Economic growth.
Qureshi. S.K., & Arif. G. M. (2001). Profile of poverty in Pakistan, 1998-99. MIMAP
Technical Paper Series, No. 5. Pakistan Institute of Development Economics
Islamabad, Pakistan.
Ram, R. (1988). Economic development and income inequality: Further evidence on the
u-curve hypothesis. World Development, 16(11), 1371–1376.
Ravallion, Martin, Chen.Shaobua. (1996). What Can New Survey Data Tell Us about
Recent Changes in Distribution and Poverty? POLICY RESEARCH WORKING
PAPER # 1694. Policy Research Department,Poverty and Human Resources
Division. The World Bank
Ravallion, Martin. (1998). Poverty Lines in Theory and Practice, Living Standards
Measurement Study Working Paper No. 133, World Bank. Washington.
206
Reyes, C. (2002). The poverty Fight: Have we made an impact, PIDS DP 2002-20 is
also of the same view, Manila
Robinson, S. (1976). A note on the u-hypothesis relating income inequality and
economic development. The American Economic Review, 66, 437-440.
Robinson, S. (1976). Sources of growth in less developed countries. Quarterly Journal
of Economics, 85(3), 391–408.
Roemer, M. and Gugerty. M. K. (1997). DOES ECONOMIC GROWTH REDUCE
POVERTY? Technical Paper, Harvard Institute for International Development.
Romer, P. (1986), increasing returns and long-run growth, Journal of Political Economy,
94, 1002-1037.
Saith, A (1983), Development and Distribution: A Critique of the Cross-Country U-
hypothesis.Journal of Development Economics No. 18.
Salz, I. S. (1999). An Examination of the Causal Relationship between Savings and
Growth in the Third World, Journal of Economics and Finance, Vol. 23: No. 1, pp
90-98.
Samuelson, P, and Modigiani. P. (1966). The Passinetti Paradox in Neo-classical and
More General Models, Review of Economic Studies, 33: pp 269-301.
Seers, D. (1969). The meaning of development. International Development Review, 11,
2-6.
207
Seers, D. (1972). What are we trying to measure. Journal of Development Studies, 8,
21-36.
Sen, A. (1983). Development: Which way now? Economic Journal, 93, 745-762.
Sen, A. (1995). Inequality reexamined. New York: Russel Sage Foundation.
Shirazi. N. S, (1993). An analysis of Pakistan’s poverty problem and its alleviation
through infaq. PhD thesis submitted to International Islamic University
Islamabad.
Shirazi. N. S, (1995). Determinants of poverty in Pakistan. Pakistan Economic and
Social Review, XXXIII (1 & 2), 91-101.
Sinha, D, and Sinha. T. (1998). Cart Before Horse? The Saving-Growth Nexus in
Mexico, Economics Letter, 61: 43-47.
Solow, R. (1956). A contribution to the theory of economic growth. Quarterly Journal of
Economics, 70(1), 65-94.
Stewart, F. (1980). Poverty and basic needs – Country experience in providing for basic
needs. The World Bank and Poverty Series.
Stewart, F. (1985). Basic needs in developing countries. Maryland: Johns Hopkins
University Press Baltimore.
Stiglitz, J. E., (2001). Globalization and its discontents. New York, Allen Lane The
Penguin Press.
208
Streeten, P. (1980). Poverty and basic needs-from growth to basic needs. The World
Bank and Poverty Series.
Swan, T. (1956) „Economic growth and capital accumulation‟, Economic Record ,
32(63), 334–361.
Tahir, S, and Ali. S.S. (1999). Growth with Equity: Policy Lessons from the experience
of Pakistan, Economic and Social Commission for Asia and Pacific Regional
Seminar on Growth with Equity. Soul.
The World Bank. (1994). Poverty Reduction in South Asia. Washington DC.
The World Bank. (1999). World Development Indicators. The World Bank, Washington
DC.
The World Bank. (2000). Global economic prospects and the developing countries.
Washington, DC: Author.
The World Bank. (2008). World development indicators 2008. Washington, DC: Author.
The World Bank. (2008). World development report 2008: Agriculture for development.
Washington, DC: Author.
The World Bank. (2011). Navigating the Strong Currents: Global Economic Prospects
Volume 2. Washington, DC: Author
Townsend, Peter. Reporting poverty in the UK A practical guide for journalists, Revised
edition 2009
209
UNDP. (2006). Poverty, Unemployment and Social Exclusion [UNDP: Croatia].
United National Conference. (February 23, 2011). On Trade and Development. Impact
of remittances on poverty in developing countries.
United Nations Development Programme. (1997). Human development report 1997.
New York: Oxford University Press.
United Nations Development Programme. (2004). Human development report 2004:
Cultural liberty in today’s diverse world. New York: Author.
Uruci, E.I. Geedeshi. (2003). Remittances Management in Albania. CeSPI working
paper 5/2003, Rome.
Weigel, V. B. (1986). The basic needs approach: Overcoming the poverty of homo
economicus. World Development, 14(12), 1423-1434.
Zaidi, S. A. (1999). Issues in Pakistan’s economy. Karachi: Oxford University Press.
Zaman, A. (1996). Statistical foundations for econometric techniques. California:
Academic Press Inc.
210
APPENDICES
Appendix I-A OLS and Empirical Bayes Estimates for RURAL PUNJAB
(Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.66** 1.09 1.49*** 1.61** 1.19** 1.48*** 1.72** 1.24** 1.51***
standard error 0.56 0.67 0.37 0.54 0.50 0.37 0.59 0.51 0.38
T- value 2.97 1.63 3.99 2.98 2.40 4.01 2.92 2.42 3.96
P- value 0.018 0.141 0.004 0.018 0.043 0.004 0.019 0.042 0.004
Ypc -0.0010** -0.0005 -0.0008** -0.001*** -0.0006* -0.0008*** -0.0010** -0.0005 -0.0008**
standard error 0.0003 0.0004 0.0002 0.0003 0.0003 0.0002 0.0003 0.0003 0.0002
T- value -3.3225 -1.3468 -3.3504 -3.4970 -2.0012 -3.5765 -3.0891 -1.8336 -3.0791
P- value 0.0105 0.2150 0.0101 0.0081 0.0804 0.0072 0.0149 0.1041 0.0151
HCI 2.704*** 1.6205* 2.0139** 2.78*** 1.47* 2.17*** 2.57** 1.31* 1.81**
standard error 0.800 0.816 0.644 0.773 0.652 0.638 0.828 0.647 0.644
T- value 3.379 1.986 3.127 3.601 2.247 3.396 3.102 2.020 2.806
P- value 0.010 0.082 0.014 0.007 0.055 0.009 0.015 0.078 0.023
B20 -0.075*** -0.062** -0.071*** -0.075*** -0.062*** -0.072*** -0.076*** -0.061*** -0.071***
standard error 0.021 0.024 0.016 0.021 0.018 0.015 0.022 0.018 0.016
T- value -3.522 -2.621 -4.570 -3.633 -3.486 -4.699 -3.381 -3.357 -4.440
P- value 0.008 0.031 0.002 0.007 0.008 0.002 0.010 0.010 0.002
Un -0.082* -0.029 -0.053 -0.081* -0.025 -0.057 -0.081* -0.026 -0.049
standard error 0.039 0.049 0.032 0.036 0.041 0.031 0.041 0.042 0.033
T- value -2.129 -0.597 -1.684 -2.226 -0.622 -1.847 -1.974 -0.610 -1.465
P- value 0.066 0.567 0.131 0.057 0.551 0.102 0.084 0.559 0.181
R Square 0.82 0.75 0.8 0.84 0.76 0.82 0.8 0.72 0.78
211
Appendix I-B OLS and Empirical Bayes Estimates for RURAL SINDH (Dependent Variable: BNGI)
Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 2.1*** 1.598** 1.84*** 2.0882*** 1.6694*** 1.8475*** 2.1077*** 1.6909*** 1.8371***
standard error 0.41 0.6290 0.3159 0.3884 0.4699 0.3067 0.4093 0.4775 0.3228
T- value 5.15 2.54 5.83 5.3762 3.5530 6.0245 5.1495 3.5408 5.6905
P- value 0.0008 0.03459 0.00030 0.0007 0.0075 0.0003 0.0009 0.0076 0.0005
Ypc -0.001*** -0.0007* -0.00096*** -0.0012*** -0.0007** -0.0010*** -0.0011*** -0.0007** -0.0009***
standard error 0.00 0.0004 0.00026 0.0003 0.0003 0.0003 0.0003 0.0003 0.0003
T- value -3.84 -1.89 -3.67 -4.1242 -2.6127 -3.9631 -3.8442 -2.5079 -3.4193
P- value 0.0049 0.09478 0.00629 0.0033 0.0310 0.0042 0.0049 0.0365 0.0091
HCI 1.44* 0.5683 1.319* 1.6146* 0.4873 1.4891** 1.4397* 0.3380 1.1415*
standard error 0.73 0.7889 0.60228 0.7107 0.6458 0.5969 0.7332 0.6259 0.5969
T- value 1.96 0.72 2.19 2.2718 0.7546 2.4947 1.9637 0.5400 1.9124
P- value 0.08510 0.49183 0.05977 0.0527 0.4721 0.0372 0.0852 0.6039 0.0922
B20 -0.058*** -0.0566** -0.053*** -0.059*** -0.0574*** -0.0547*** -0.0580*** -0.0561*** -0.0517***
standard error 0.01 0.0214 0.01238 0.0124 0.0162 0.0118 0.0131 0.0164 0.0129
T- value -4.44 -2.65 -4.28 -4.7662 -3.5486 -4.6185 -4.4352 -3.4138 -4.0138
P- value 0.0022 0.02948 0.00267 0.0014 0.0075 0.0017 0.0022 0.0092 0.0039
Un -0.19* -0.0448 -0.169* -0.2045** -0.0388 -0.1813* -0.1944* -0.0325 -0.1583
standard error 0.09 0.0574 0.08312 0.0829 0.0501 0.0796 0.0868 0.0502 0.0861
T- value -2.24 -0.78 -2.04 -2.4668 -0.7738 -2.2764 -2.2393 -0.6472 -1.8391
P- value 0.0550 0.45720 0.07516 0.0389 0.4613 0.0524 0.0555 0.5356 0.1032
R Square 0.81 0.65 0.81 0.84 0.67 0.83 0.8 0.65 0.79
212
Appendix I-C OLS and Empirical Bayes Estimates for RURAL KPK (Dependent Variable: BNGI)
Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 2.7967** 1.949** 1.5869*** 2.8114** 1.7208** 1.5780*** 2.7967** 1.7399** 1.5955***
standard error 0.9367 0.7524 0.4349 0.9519 0.5508 0.4381 0.9367 0.5587 0.4298
T- value 2.9857 2.5914 3.6486 2.9535 3.1244 3.6015 2.9857 3.1143 3.7119
P- value 0.0174 0.0320 0.0065 0.0183 0.0141 0.0070 0.0174 0.0144 0.0059
Ypc -0.0010* -0.0006 -0.0006 -0.0010* -0.0005 -0.0006 -0.0010* -0.0005 -0.0006
standard error 0.0005 0.0004 0.0003 0.0005 0.0003 0.0003 0.0005 0.0003 0.0003
T- value -2.1012 -1.4362 -1.8154 -2.0658 -1.6524 -1.7823 -2.1012 -1.6731 -1.8564
P- value 0.0688 0.1889 0.1070 0.0727 0.1371 0.1126 0.0688 0.1328 0.1005
HCI -1.4400 -1.1696 -0.4169 -1.4875 -0.8916 -0.4008 -1.4406 -0.8825 -0.4266
standard error 0.7864 0.8020 0.5425 0.8252 0.6491 0.5648 0.7864 0.6271 0.5196
T- value -1.8318 -1.4582 -0.7680 -1.8025 -1.3736 -0.7097 -1.8319 -1.4073 -0.8210
P- value 0.1043 0.1828 0.4642 0.1091 0.2068 0.4981 0.1043 0.1970 0.4354
B20 -0.0603* -0.049* -0.0276 -0.0604 -0.0471** -0.0274 -0.0604* -0.0483** -0.0279
standard error 0.0322 0.0258 0.0222 0.0325 0.0193 0.0223 0.0323 0.0196 0.0221
T- value -1.8710 -1.8970 -1.2400 -1.8571 -2.4419 -1.2249 -1.8710 -2.4690 -1.2617
P- value 0.0982 0.0943 0.2494 0.1004 0.0404 0.2555 0.0983 0.0388 0.2426
Un 0.0205 0.0330 -0.0220 0.0206 0.0267 -0.0232 0.0205 0.0268 -0.0215
standard error 0.0517 0.0519 0.0451 0.0522 0.0434 0.0453 0.0517 0.0437 0.0449
T- value 0.3970 0.6408 -0.4939 0.3953 0.6161 -0.5112 0.3970 0.6139 -0.4801
P- value 0.7017 0.5396 0.6346 0.7029 0.5550 0.6230 0.7017 0.5564 0.6440
R Square 0.51 0.35 0.39 0.51 0.29 0.38 0.51 0.3 0.4
213
Appendix I-D OLS and Empirical Bayes Estimates for RURAL BALOCHISTAN (Dependent Variable: BNGI)
Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.93*** 1.4821* 1.6460*** 1.93*** 1.5063** 1.6383*** 1.9321*** 1.5017** 1.6535***
standard error 0.4950 0.6490 0.3500 0.5012 0.4856 0.3547 0.4960 0.4923 0.3493
T- value 3.8954 2.2800 4.6710 3.8520 3.1017 4.6187 3.8955 3.0503 4.7333
P- value 0.0045 0.0521 0.0016 0.0049 0.0146 0.0017 0.0046 0.0158 0.0015
Ypc -0.0010** -0.0006 -0.0008*** -0.0010** -0.0006* -0.0008*** -0.0010** -0.0006* -0.0008***
standard error 0.0003 0.0004 0.0002 0.0003 0.0003 0.0002 0.0003 0.0003 0.0002
T- value -3.2165 -1.6618 -3.5346 -3.2069 -2.2232 -3.5237 -3.2165 -2.2001 -3.5522
P- value 0.0123 0.1351 0.0077 0.0125 0.0569 0.0078 0.0123 0.0590 0.0075
HCI 0.0180 -0.1050 0.2920 0.0239 0.0002 0.3306 0.0189 -0.0320 0.2533
standard error 0.6928 0.7941 0.5788 0.7163 0.6338 0.5977 0.6928 0.6115 0.5558
T- value 0.0273 -0.1330 0.5047 0.0333 0.0003 0.5532 0.0273 -0.0524 0.4557
P- value 0.9789 0.8975 0.6274 0.9742 0.9998 0.5953 0.9789 0.9595 0.6607
B20 -0.0470** -0.0440* -0.0454*** -0.0471** -0.0469** -0.046*** -0.0470** -0.0453** -0.0448***
standard error 0.0153 0.0219 0.0133 0.0155 0.0166 0.0136 0.0153 0.0168 0.0130
T- value -3.0675 -2.0078 -3.4040 -3.0337 -2.8163 -3.3797 -3.0675 -2.7014 -3.4394
P- value 0.0154 0.0796 0.0093 0.0162 0.0226 0.0096 0.0154 0.0270 0.0088
Un 0.0325 0.0401 0.0266 0.0323 0.0326 0.0253 0.0325 0.0336 0.0280
standard error 0.0353 0.0489 0.0328 0.0357 0.0413 0.0332 0.0353 0.0415 0.0324
T- value 0.9208 0.8210 0.8115 0.9061 0.7906 0.7632 0.9208 0.8109 0.8635
P- value 0.3841 0.4354 0.4405 0.3913 0.4520 0.4673 0.3841 0.4409 0.4130
R Square 0.72 0.63 0.71 0.72 0.63 0.71 0.72 0.63 0.71
214
Appendix II-A OLS and Empirical Bayes Estimates for URBAN PUNJAB (Dependent Variable: BNGI)
Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.59*** 1.41** 1.33*** 1.59*** 1.3629** 1.3013*** 1.5935*** 1.4020*** 1.3645***
standard error 0.3890 0.4299 0.1783 0.41 0.4747 0.1287 0.3756 0.4119 0.2058
T- value 4.10 3.28 7.51 3.91 2.8713 10.1147 4.2423 3.4035 6.6295
P- value 0.0030 0.0110 0.0001 0.0046 0.0208 0.0000 0.0028 0.0093 0.0002
Ypc -0.0002 -0.00012 -0.0001 0.00 -0.0001 -0.0001 -0.0002 -0.0001 -0.0001
standard error 0.0002 0.0002 0.0001 0.00 0.0002 0.0001 0.0002 0.0002 0.0001
T- value -0.90 -0.59 -1.43 -1.01 -0.6440 -1.5042 -0.8221 -0.7351 -1.3208
P- value 0.3900 0.5708 0.1896 0.3431 0.5376 0.1710 0.4348 0.4833 0.2231
HCI -0.31 -0.25 -0.18** -0.22 -0.1948 -0.1564 -0.3686 -0.2522 -0.1928
standard error 0.7649 0.6191 0.0603 0.81 0.7084 0.1230 0.7231 0.5767 0.1105
T- value -0.40 -0.40 -2.96 -0.28 -0.2750 -1.2717 -0.5097 -0.4373 -1.7450
P- value 0.7004 0.7009 0.0183 0.7889 0.7903 0.2392 0.6240 0.6734 0.1191
B20 -0.08*** -0.07*** -0.07*** -0.08*** -0.0726*** -0.0730*** -0.0805*** -0.0738*** -0.0747***
standard error 0.0118 0.0111 0.0075 0.01 0.0119 0.0067 0.0118 0.0110 0.0081
T- value -6.82 -6.68 -9.83 -6.76 -6.0798 -10.9534 -6.8400 -6.7233 -9.2450
P- value 0.0001 0.0002 0.0000 0.0001 0.0003 0.0000 0.0001 0.0001 0.0000
Un 0.011 0.003 0.010 0.01 0.0022 0.0098 0.0125 0.0035 0.0104
standard error 0.0148 0.0169 0.0109 0.01 0.0178 0.0110 0.0147 0.0167 0.0109
T- value 0.77 0.20 0.93 0.69 0.1249 0.8887 0.8503 0.2121 0.9552
P- value 0.4621 0.8445 0.3794 0.5106 0.9037 0.4001 0.4199 0.8373 0.3675
R Square 0.93 0.92 0.92 0.92 0.86 0.92 0.93 0.83 0.92
215
Appendix II-B OLS and Empirical Bayes Estimates for URBAN SINDH (Dependent Variable: BNGI)
Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.54** 1.29** 1.30*** 1.5111** 1.2460** 1.2771*** 1.5474** 1.3139** 1.3080***
standard error 0.5665 0.4545 0.1892 0.5995 0.4962 0.1325 0.5399 0.4276 0.2232
T- value 2.72 2.83 6.85 2.5206 2.5111 9.6349 2.8659 3.0731 5.8592
P- value 0.0264 0.0221 0.0001 0.0358 0.0363 0.0000 0.0210 0.0153 0.0004
Ypc -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001 -0.0001
standard error 0.0002 0.0002 0.0001 0.0002 0.0002 0.0001 0.0002 0.0002 0.0001
T- value -0.51 -0.28 -1.04 -0.6147 -0.3755 -1.1142 -0.4227 -0.4188 -0.9211
P- value 0.6237 0.7833 0.3286 0.5559 0.7171 0.2976 0.6837 0.6864 0.3839
HCI -0.39 -0.21 -0.18** -0.2740 -0.1655 -0.1597 -0.4637 -0.2724 -0.1947
standard error 0.8363 0.6438 0.0603 0.9121 0.7239 0.1232 0.7670 0.5850 0.1105
T- value -0.46 -0.32 -2.97 -0.3004 -0.2286 -1.2966 -0.6046 -0.4656 -1.7619
P- value 0.6553 0.7581 0.0180 0.7715 0.8249 0.2309 0.5622 0.6539 0.1161
B20 -0.07*** -0.07*** -0.07*** -0.071*** -0.067*** -0.066*** -0.07*** -0.0681*** -0.067***
standard error 0.0121 0.0112 0.0074 0.0123 0.0120 0.0069 0.0119 0.0108 0.0077
T- value -5.87 -6.05 -9.09 -5.7853 -5.5925 -9.6955 -5.9256 -6.2755 -8.6874
P- value 0.0004 0.0003 0.0000 0.0004 0.0005 0.0000 0.0004 0.0002 0.0000
Un 0.0008 -0.0010 0.0081 0.0010 -0.0012 0.0087 0.0008 -0.0008 0.0075
standard error 0.0327 0.0200 0.0275 0.0330 0.0210 0.0272 0.0324 0.0195 0.0276
T- value 0.02 -0.05 0.29 0.0301 -0.0578 0.3213 0.0242 -0.0389 0.2725
P- value 0.9807 0.9602 0.7767 0.9768 0.9553 0.7562 0.9813 0.9700 0.7921
R Square 0.91 0.91 0.91 0.91 0.88 0.91 0.91 0.87 0.91
216
Appendix II-C OLS and Empirical Bayes Estimates for URBAN KPK (Dependent Variable: BNGI)
Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 0.44 0.97* 1.18*** 0.5333 1.1092* 1.2199*** 0.3530 0.9479* 1.1450***
standard error 0.7678 0.4993 0.1881 0.7864 0.5529 0.1323 0.7453 0.4772 0.2221
T- value 0.5700 1.9427 6.2597 0.6781 2.0061 9.2235 0.4736 1.9865 5.1557
P- value 0.5843 0.0880 0.0002 0.5168 0.0798 0.0000 0.6484 0.0822 0.0009
Ypc 0.0002 0.00003 -0.0001 0.0002 0.0000 -0.0001 0.0002 0.0000 0.0000
standard error 0.0002 0.0002 0.0001 0.0002 0.0002 0.0001 0.0002 0.0002 0.0001
T- value 0.78 0.16 -0.51 0.6795 -0.0950 -0.7418 0.8659 -0.0045 -0.3433
P- value 0.4601 0.8775 0.6232 0.5160 0.9267 0.4794 0.4118 0.9965 0.7402
HCI 0.14 -0.11 -0.18** 0.0262 -0.2502 -0.1855 0.2519 -0.0810 -0.1999
standard error 0.7348 0.6061 0.0600 0.7450 0.6886 0.1205 0.7187 0.5701 0.1090
T- value 0.19 -0.18 -3.04 0.0352 -0.3633 -1.5400 0.3506 -0.1421 -1.8331
P- value 0.8504 0.8604 0.0160 0.9728 0.7258 0.1621 0.7350 0.8905 0.1041
B20 -0.038* -0.056*** -0.0572*** -0.039* -0.059*** -0.058*** -0.0383* -0.0555*** -0.055***
standard error 0.0186 0.0124 0.0100 0.0189 0.0134 0.0092 0.0183 0.0122 0.0108
T- value -2.09 -4.50 -5.71 -2.1022 -4.4278 -6.3984 -2.0894 -4.5535 -5.1483
P- value 0.0703 0.0020 0.0004 0.0687 0.0022 0.0002 0.0701 0.0019 0.0009
Un 0.0029 -0.0001 -0.0002 0.0043 0.0012 -0.0008 0.0015 -0.0003 0.0007
standard error 0.0152 0.0169 0.0105 0.0151 0.0179 0.0106 0.0152 0.0167 0.0107
T- value 0.19 0.00 -0.02 0.2867 0.0664 -0.0730 0.0971 -0.0180 0.0684
P- value 0.8534 0.9976 0.9858 0.7816 0.9487 0.9436 0.9250 0.9860 0.9471
R Square 0.79 0.76 0.76 0.79 0.65 0.75 0.79 0.61 0.76
217
Appendix II-D OLS and Empirical Bayes Estimates for URBAN BALOCHISTAN (Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.45** 1.24** 1.32*** 1.42** 1.229** 1.2861*** 1.4803** 1.2682** 1.3338***
standard error 0.5845 0.4643 0.1863 0.61 0.4969 0.1315 0.5592 0.4447 0.2191
T- value 2.49 2.66 7.06 2.31 2.4733 9.7772 2.6470 2.8518 6.0883
P- value 0.0377 0.0287 0.0001 0.049 0.0385 0.0000 0.0294 0.0214 0.0003
Ypc -0.0004 -0.0001 -0.0002 0.00 -0.0002 -0.0002 -0.0004 -0.0002 -0.0002
standard error 0.0003 0.0002 0.0002 0.00 0.0002 0.0002 0.0003 0.0002 0.0002
T- value -1.13 -0.63 -1.15 -1.13 -0.7153 -1.2267 -1.1299 -0.7771 -1.1156
P- value 0.2909 0.5444 0.2821 0.2926 0.4948 0.2548 0.2913 0.4594 0.2970
HCI -0.0127 -0.0720 -0.1744** 0.05 -0.0831 -0.1429 -0.0593 -0.1356 -0.1794
standard error 0.6253 0.5938 0.0600 0.66 0.6559 0.1210 0.5882 0.5560 0.1091
T- value -0.02 -0.12 -2.91 0.07 -0.1267 -1.1810 -0.1009 -0.2439 -1.6445
P- value 0.9843 0.9065 0.0197 0.9455 0.9023 0.2715 0.9221 0.8134 0.1387
B20 -0.058*** -0.063*** -0.0591*** -0.06*** -0.0634*** -0.0589*** -0.0590*** -0.0643*** -0.0592***
standard error 0.0125 0.0112 0.0116 0.01 0.0117 0.0116 0.0124 0.0110 0.0116
T- value -4.68 -5.69 -5.10 -4.59 -5.4272 -5.0726 -4.7694 -5.8278 -5.0958
P- value 0.0016 0.0005 0.0009 0.0017 0.0006 0.0010 0.0014 0.0004 0.0009
Un 0.0191 0.0051 0.0179 0.02 0.0050 0.0168 0.0201 0.0059 0.0186
standard error 0.0325 0.0200 0.0291 0.03 0.0203 0.0290 0.0325 0.0199 0.0292
T- value 0.59 0.25 0.61 0.55 0.2465 0.5801 0.6177 0.2966 0.6374
P- value 0.5737 0.8066 0.5558 0.5955 0.8115 0.5778 0.5539 0.7744 0.5417
R Square 0.84 0.83 0.83 0.84 0.81 0.83 0.84 0.79 0.83
218
Appendix III-A OLS and Empirical Bayes Estimates for OVERALL PUNJAB (Dependent Variable: BNGI)
Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.30** 1.31** 1.61*** 1.145* 1.298** 1.579*** 1.4585** 1.3724** 1.636***
standard error 0.5160 0.4781 0.1353 0.5128 0.4799 0.1602 0.5301 0.4556 0.1227
T- value 2.52 2.74 11.91 2.2329 2.7048 9.8611 2.7511 3.0125 13.3358
P- value 0.03556 0.02541 0.000002 0.0560 0.0269 0.0000 0.0250 0.0167 0.0000
Ypc -0.0015*** -0.0007** -0.0012*** -0.0014*** -0.0007** -0.0012*** -0.0015*** -0.0007** -0.0011***
standard error 0.0004 0.0003 0.0003 0.0004 0.0003 0.0003 0.0004 0.0003 0.0003
T- value -3.94 -2.41 -4.43 -4.0469 -2.3148 -4.5038 -3.7569 -2.3310 -4.2738
P- value 0.0043 0.0428 0.0022 0.0037 0.0493 0.0020 0.0056 0.0481 0.0027
HCI 4.14*** 1.44 2.60** 4.1798*** 1.3036 2.6737** 4.0217** 1.5576* 2.4365**
standard error 1.2103 0.8709 0.9107 1.1827 0.8594 0.8984 1.2419 0.7934 0.9139
T- value 3.42 1.66 2.86 3.5340 1.5169 2.9760 3.2383 1.9632 2.6660
P- value 0.0090 0.1360 0.0212 0.0077 0.1678 0.0177 0.0119 0.0852 0.0285
B20 -0.07** -0.06*** -0.08*** -0.0617** -0.0626*** -0.0747*** -0.0712** -0.0648*** -0.0749***
standard error 0.0214 0.0174 0.0094 0.0210 0.0174 0.0099 0.0222 0.0167 0.0093
T- value -3.12 -3.58 -8.00 -2.9453 -3.6002 -7.5812 -3.2121 -3.8738 -8.0671
P- value 0.0143 0.0072 0.00004 0.0186 0.0070 0.0001 0.0124 0.0047 0.0000
Un -0.07* -0.03 -0.04 -0.0638* -0.0232 -0.0388 -0.0711* -0.0292 -0.0381
standard error 0.0317 0.0289 0.0260 0.0297 0.0277 0.0251 0.0339 0.0276 0.0269
T- value -2.16 -0.95 -1.52 -2.1447 -0.8398 -1.5429 -2.0991 -1.0597 -1.4128
P- value 0.0625 0.3692 0.1665 0.0643 0.4254 0.1614 0.0690 0.3202 0.1954
R Square 0.77 0.32 0.71 0.78 0.33 0.72 0.75 0.35 0.7
219
Appendix III-B OLS and Empirical Bayes Estimates for OVERALL SINDH (Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 2.06*** 1.69*** 1.66*** 1.946*** 1.6313*** 1.6544*** 2.1635*** 1.7365*** 1.6718***
standard error 0.5432 0.4692 0.1356 0.5106 0.4531 0.1600 0.5655 0.4594 0.1230
T- value 3.80 3.59 12.24 3.8126 3.6002 10.3433 3.8254 3.7801 13.5925
P- value 0.0052 0.0071 0.000002 0.0051 0.0070 0.0000 0.0051 0.0054 0.0000
Ypc -0.0009*** -0.0006* -0.0008*** -0.0010*** -0.0007** -0.0009*** -0.0009** -0.0005 -0.0007**
standard error 0.0003 0.0003 0.0002 0.0002 0.0003 0.0002 0.0003 0.0003 0.0002
T- value -3.51 -2.25 -3.61 -3.9801 -2.5859 -4.1190 -3.1691 -1.7413 -3.2355
P- value 0.0080 0.0544 0.0069 0.0041 0.0323 0.0033 0.0132 0.1198 0.0120
HCI 1.36 0.68 1.38* 1.6835* 0.9416 1.6477** 1.0661 0.4821 1.1260
standard error 0.7564 0.7374 0.6299 0.7376 0.7176 0.6121 0.7505 0.7018 0.6292
T- value 1.80 0.92 2.19 2.2824 1.3122 2.6918 1.4206 0.6870 1.7895
P- value 0.1099 0.3865 0.0598 0.0519 0.2259 0.0274 0.1932 0.5115 0.1113
B20 -0.09*** -0.07*** -0.07*** -0.083*** -0.073*** -0.074*** -0.088*** -0.074*** -0.073***
standard error 0.0197 0.0165 0.0117 0.0183 0.0158 0.0112 0.0208 0.0164 0.0122
T- value -4.35 -4.47 -6.26 -4.5404 -4.6180 -6.6009 -4.2433 -4.5468 -5.9877
P- value 0.0024 0.0021 0.0002 0.0019 0.0017 0.0002 0.0028 0.0019 0.0003
Un -0.05 -0.03 -0.04 -0.0551 -0.0330 -0.0438 -0.0457 -0.0256 -0.0301
standard error 0.0386 0.0289 0.0360 0.0357 0.0274 0.0336 0.0407 0.0288 0.0376
T- value -1.31 -0.97 -1.02 -1.5464 -1.2068 -1.3057 -1.1227 -0.8891 -0.8010
P- value 0.2279 0.3591 0.3371 0.1606 0.2620 0.2279 0.2941 0.3999 0.4462
R Square 0.8 0.58 0.78 0.83 0.65 0.82 0.78 0.5 0.75
220
Appendix III-C OLS and Empirical Bayes Estimates for OVERALL KPK (Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 2.10** 1.71** 1.6*** 2.102** 1.7301*** 1.6212*** 2.1005** 1.6787*** 1.6490***
standard error 0.8357 0.5140 0.1380 0.8408 0.5044 0.1650 0.8308 0.4981 0.1244
T- value 2.51 3.32 11.84 2.5002 3.4303 9.8276 2.5285 3.3702 13.2503
P- value 0.0361 0.0105 0.000002 0.0369 0.0090 0.0000 0.0353 0.0098 0.0000
Ypc -0.0003 -0.0001 -0.0001 -0.0003 -0.0001 -0.0001 -0.0003 0.0000 -0.0001
standard error 0.0004 0.0003 0.0002 0.0004 0.0003 0.0002 0.0004 0.0003 0.0002
T- value -0.71 -0.51 -0.66 -0.6828 -0.5374 -0.5995 -0.7286 -0.0606 -0.7258
P- value 0.5000 0.6230 0.5294 0.5140 0.6056 0.5654 0.4870 0.9532 0.4887
HCI -1.04 -1.02 -0.54 -1.0623 -1.0446 -0.5331 -1.0273 -1.0930 -0.5516
standard error 0.7906 0.7344 0.5018 0.8111 0.7229 0.5153 0.7710 0.6879 0.4895
T- value -1.32 -1.39 -1.09 -1.3097 -1.4451 -1.0345 -1.3324 -1.5889 -1.1269
P- value 0.2229 0.2007 0.3095 0.2267 0.1864 0.3312 0.2194 0.1507 0.2925
B20 -0.08*** -0.07*** -0.07*** -0.0793*** -0.0709*** -0.0703*** -0.0796*** -0.0689*** -0.0710***
standard error 0.0235 0.0169 0.0149 0.0235 0.0165 0.0151 0.0235 0.0166 0.0148
T- value -3.38 -4.18 -4.73 -3.3703 -4.3092 -4.6528 -3.3887 -4.1579 -4.7852
P- value 0.0096 0.0031 0.0015 0.0098 0.0026 0.0016 0.0095 0.0032 0.0014
Un 0.0098 0.0119 -0.0040 0.0098 0.0117 -0.0045 0.0098 0.0133 -0.0035
standard error 0.0403 0.0292 0.0359 0.0404 0.0282 0.0360 0.0402 0.0286 0.0358
T- value 0.24 0.41 -0.11 0.2415 0.4142 -0.1259 0.2448 0.4668 -0.0981
P- value 0.8138 0.6937 0.9143 0.8153 0.6896 0.9029 0.8128 0.6531 0.9243
R Square 0.66 0.32 0.64 0.66 0.33 0.64 0.66 0.63 0.64
221
Appendix III-D OLS and Empirical Bayes Estimates for OVERALL BALOCHISTAN (Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 2.10*** 1.72*** 1.67*** 2.089*** 1.7105*** 1.6705*** 2.114*** 1.7060*** 1.6794***
standard error 0.4575 0.4532 0.1338 0.4616 0.4432 0.1581 0.4530 0.4424 0.1213
T- value 4.59 3.80 12.48 4.5269 3.8596 10.5651 4.6674 3.8560 13.8448
P- value 0.0018 0.0052 0.0000 0.0019 0.0048 0.0000 0.0016 0.0048 0.0000
Ypc -0.0009** -0.0006* -0.0007*** -0.0009** -0.0005* -0.0007*** -0.0009** -0.0004 -0.0007***
standard error 0.0003 0.0003 0.0001 0.0003 0.0003 0.0001 0.0003 0.0003 0.0001
T- value -3.19 -2.09 -4.75 -3.1663 -2.1639 -4.5183 -3.2241 -1.6684 -4.8855
P- value 0.0127 0.0700 0.0015 0.0133 0.0624 0.0020 0.0122 0.1338 0.0012
HCI 0.41 0.09 0.69 0.4279 0.1158 0.7056 0.3935 0.0521 0.6541
standard error 0.5334 0.6938 0.4311 0.5393 0.6758 0.4351 0.5246 0.6554 0.4265
T- value 0.77 0.13 1.59 0.7935 0.1714 1.6216 0.7500 0.0795 1.5338
P- value 0.4612 0.9008 0.1505 0.4504 0.8682 0.1435 0.4747 0.9386 0.1636
B20 -0.069*** -0.067*** -0.066*** -0.069*** -0.067*** -0.0671*** -0.0688*** -0.0654*** -0.0657***
standard error 0.0121 0.0152 0.0115 0.0121 0.0148 0.0117 0.0120 0.0150 0.0114
T- value -5.72 -4.39 -5.76 -5.7113 -4.5341 -5.7530 -5.7357 -4.3737 -5.7728
P- value 0.0004 0.0023 0.0004 0.0004 0.0019 0.0004 0.0004 0.0024 0.0004
Un 0.0050 -0.0007 -0.0032 0.0046 -0.0018 -0.0037 0.0055 -0.0001 -0.0023
standard error 0.0282 0.0275 0.0263 0.0281 0.0264 0.0262 0.0282 0.0271 0.0263
T- value 0.18 -0.03 -0.12 0.1647 -0.0685 -0.1421 0.1949 -0.0030 -0.0872
P- value 0.8644 0.9802 0.9057 0.8733 0.9471 0.8905 0.8503 0.9977 0.9327
R Square 0.86 0.69 0.85 0.86 0.7 0.85 0.86 0.7 0.84
222
Appendix IV-A OLS and Empirical Bayes Estimates for RURAL PUNJAB (Dependent Variable: BNGI)
Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.67** 1.16 1.49*** 1.6053** 1.1354** 1.5443** 1.7189** 1.2169** 1.5061***
standard error 0.5601 0.7186 0.4017 0.5382 0.4323 0.5094 0.5893 0.4333 0.4171
T- value 2.97 1.61 3.71 2.9825 2.6264 3.0318 2.9171 2.8088 3.6109
P- value 0.0178 0.1457 0.0060 0.0175 0.0303 0.0163 0.0194 0.0229 0.0069
Ypc -0.0011** -0.0005 -0.0008** -0.0011*** -0.0004 -0.0007** -0.0010** -0.0004 -0.0007**
standard error 0.0003 0.0004 0.0003 0.0003 0.0002 0.0003 0.0003 0.0002 0.0003
T- value -3.32 -1.49 -3.07 -3.4970 -1.8488 -2.7540 -3.0891 -1.6987 -2.7835
P- value 0.0105 0.1747 0.0153 0.0081 0.1017 0.0249 0.0149 0.1278 0.0238
HCI 2.70*** 1.95* 1.88** 2.7830*** 1.3893** 1.5668** 2.5677** 1.1500* 1.6746**
standard error 0.8002 0.9060 0.6589 0.7728 0.6019 0.5634 0.8277 0.5746 0.6624
T- value 3.38 2.15 2.86 3.6012 2.3083 2.7811 3.1023 2.0014 2.5282
P- value 0.0097 0.0638 0.0213 0.0070 0.0498 0.0239 0.0146 0.0803 0.0354
B20 -0.075*** -0.0675** -0.0714*** -0.0749*** -0.0632*** -0.0751*** -0.0760*** -0.0642*** -0.0714***
standard error 0.0214 0.0230 0.0164 0.0206 0.0147 0.0197 0.0225 0.0150 0.0171
T- value -3.52 -2.93 -4.35 -3.6328 -4.3076 -3.8188 -3.3808 -4.2692 -4.1770
P- value 0.0078 0.0189 0.0024 0.0067 0.0026 0.0051 0.0096 0.0027 0.0031
Un -0.0820* -0.0389 -0.0494 -0.0811* -0.0261 -0.0399 -0.0807* -0.0224 -0.0441
standard error 0.0385 0.0504 0.0328 0.0364 0.0285 0.0314 0.0409 0.0294 0.0341
T- value -2.13 -0.77 -1.51 -2.2258 -0.9153 -1.2702 -1.9737 -0.7624 -1.2940
P- value 0.0658 0.4630 0.1697 0.0567 0.3868 0.2397 0.0839 0.4677 0.2318
R Square 0.82 0.65 0.8 0.84 0.62 0.78 0.8 0.64 0.77
223
Appendix IV-B OLS and Empirical Bayes Estimates for URBAN PUNJAB (Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.60*** 1.25* 1.50*** 1.5893*** 1.2668** 1.5337*** 1.5935*** 1.3305*** 1.5093***
standard error 0.3890 0.6572 0.3181 0.4061 0.4328 0.3764 0.3756 0.3819 0.3144
T- value 4.10 1.90 4.71 3.9133 2.9271 4.0743 4.2423 3.4843 4.8005
P- value 0.0034 0.0940 0.0015 0.0045 0.0191 0.0036 0.0028 0.0083 0.0014
Ypc -0.0002 0.0001 -0.0002 -0.0002 -0.0001 -0.0003 -0.0002 -0.0001 -0.0002
standard error 0.0002 0.0003 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002
T- value -0.91 0.19 -1.07 -1.0076 -0.3000 -1.3384 -0.8221 -0.2624 -0.9875
P- value 0.3900 0.8544 0.3146 0.3431 0.7718 0.2176 0.4348 0.7997 0.3523
HCI -0.31 -0.34 -0.13 -0.2233 -0.0337 -0.0180 -0.3686 -0.2050 -0.2018
standard error 0.7649 0.9066 0.6298 0.8067 0.6841 0.5764 0.7231 0.5646 0.6008
T- value -0.40 -0.37 -0.20 -0.2768 -0.0492 -0.0311 -0.5097 -0.3630 -0.3359
P- value 0.7004 0.7210 0.8471 0.7889 0.9619 0.9759 0.6240 0.7260 0.7456
B20 -0.081*** -0.077*** -0.079*** -0.081*** -0.0705*** -0.0803*** -0.0805*** -0.0725*** -0.0788***
standard error 0.0118 0.0187 0.0104 0.0120 0.0130 0.0117 0.0118 0.0119 0.0105
T- value -6.82 -4.12 -7.55 -6.7600 -5.4305 -6.8754 -6.8400 -6.0787 -7.5220
P- value 0.0001 0.0033 0.0001 0.0001 0.0006 0.0001 0.0001 0.0003 0.0001
Un 0.011 0.023 0.009 0.0102 0.0075 0.0077 0.0125 0.0096 0.0103
standard error 0.0148 0.0433 0.0137 0.0148 0.0227 0.0131 0.0147 0.0216 0.0137
T- value 0.77 0.54 0.67 0.6884 0.3289 0.5857 0.8503 0.4443 0.7553
P- value 0.4621 0.6059 0.5215 0.5107 0.7507 0.5742 0.4199 0.6686 0.4717
R Square 0.93 0.23 0.93 0.93 0.44 0.92 0.93 0.6 0.93
224
Appendix IV-C OLS and Empirical Bayes Estimates for RURAL SINDH
(Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
2.11*** 1.62** 1.89*** 2.0882*** 1.468*** 2.1118*** 2.1302*** 1.536*** 1.8985***
standard error 0.4093 0.6627 0.3317 0.3884 0.4039 0.3720 0.4280 0.3922 0.3445
T- value 5.15 2.45 5.69 5.3762 3.6345 5.6771 4.9772 3.9182 5.5113
P- value 0.0009 0.0399 0.0005 0.0007 0.0066 0.0005 0.0011 0.0044 0.0006 Ypc -0.0011*** -0.0007* -0.0010*** -0.0012*** -0.0005* -0.0010*** -0.0011*** -0.0005* -0.0009**
standard error 0.0003 0.0004 0.0003 0.0003 0.0002 0.0003 0.0003 0.0002 0.0003
T- value -3.84 -1.89 -3.56 -4.1242 -2.1518 -3.8741 -3.6015 -2.1679 -3.3180
P- value 0.0049 0.0956 0.0074 0.0033 0.0636 0.0047 0.0070 0.0620 0.0106
HCI 1.44* 0.98 1.17* 1.6146* 0.6620 1.0017* 1.2510 0.4962 0.9823
standard error 0.7332 0.8950 0.6154 0.7107 0.6232 0.5384 0.7435 0.5808 0.6128
T- value 1.96 1.09 1.90 2.2718 1.0622 1.8604 1.6826 0.8542 1.6031
P- value 0.0852 0.3074 0.0944 0.0527 0.3191 0.0999 0.1310 0.4178 0.1476
B20 -0.058*** -0.059** -0.053*** -0.059*** -0.061*** -0.058*** -0.057*** -0.0608*** -0.0524***
standard error 0.0131 0.0192 0.0125 0.0124 0.0124 0.0123 0.0137 0.0123 0.0130
T- value -4.44 -3.05 -4.27 -4.7662 -4.9154 -4.7087 -4.1661 -4.9291 -4.0192
P- value 0.0022 0.0157 0.0027 0.0014 0.0012 0.0015 0.0031 0.0012 0.0038
Un -0.19* -0.10 -0.16* -0.2045** -0.0305 -0.1657* -0.1837* -0.0263 -0.1526
standard error 0.0868 0.0721 0.0836 0.0829 0.0383 0.0780 0.0901 0.0381 0.0867
T- value -2.24 -1.40 -1.96 -2.4668 -0.7977 -2.1251 -2.0387 -0.6907 -1.7606
P- value 0.0555 0.1993 0.0860 0.0389 0.4480 0.0663 0.0758 0.5093 0.1163
R Square 0.82 0.7 0.81 0.84 0.62 0.82 0.8 0.62 0.79
225
Appendix IV-D OLS and Empirical Bayes Estimates for URBAN SINDH (Dependent Variable: BNGI)
Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.54** 1.15 1.41*** 1.5111** 1.1469** 1.4125** 1.5474** 1.2023** 1.4292***
standard error 0.5665 0.7193 0.3954 0.5995 0.4570 0.5207 0.5399 0.4021 0.3924
T- value 2.72 1.60 3.57 2.5206 2.5094 2.7129 2.8659 2.9901 3.6422
P- value 0.0264 0.1479 0.0073 0.0358 0.0364 0.0265 0.0210 0.0173 0.0066
Ypc -0.0001 0.0001 -0.0001 -0.0001 0.0000 -0.0002 -0.0001 0.0000 -0.0001
standard error 0.0002 0.0003 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002
T- value -0.51 0.40 -0.69 -0.6147 -0.0510 -0.9095 -0.4227 0.0668 -0.5929
P- value 0.6237 0.6983 0.5114 0.5559 0.9606 0.3897 0.6837 0.9483 0.5696
HCI -0.39 -0.33 -0.16 -0.2740 -0.0156 -0.0158 -0.4637 -0.1963 -0.2516
standard error 0.8363 0.9385 0.6515 0.9121 0.6928 0.6091 0.7670 0.5675 0.6132
T- value -0.46 -0.36 -0.25 -0.3004 -0.0225 -0.0259 -0.6046 -0.3458 -0.4103
P- value 0.6553 0.7307 0.8071 0.7715 0.9826 0.9800 0.5622 0.7384 0.6923
B20 -0.071*** -0.068*** -0.070*** -0.0710*** -0.0633*** -0.0704*** -0.0708*** -0.0642*** -0.0695***
standard error 0.0121 0.0187 0.0100 0.0123 0.0129 0.0118 0.0119 0.0115 0.0101
T- value -5.87 -3.62 -6.92 -5.7853 -4.8863 -5.9736 -5.9256 -5.5680 -6.9077
P- value 0.0004 0.0068 0.0001 0.0004 0.0012 0.0003 0.0004 0.0005 0.0001
Un 0.0008 0.0155 0.0034 0.0010 0.0085 0.0019 0.0008 0.0087 0.0034
standard error 0.0327 0.0493 0.0303 0.0330 0.0319 0.0324 0.0324 0.0281 0.0301
T- value 0.0249 0.3153 0.1121 0.0301 0.2656 0.0576 0.0242 0.3085 0.1121
P- value 0.9807 0.7606 0.9135 0.9768 0.7972 0.9555 0.9813 0.7656 0.9135
R Square 0.91 0.43 0.91 0.91 0.6 0.91 0.91 0.68 0.91
226
Appendix IV-E OLS and Empirical Bayes Estimates for RURAL KPK (Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 2.80** 2.04** 1.66*** 2.8114** 1.567** 1.9650** 2.7823** 1.554** 1.6854***
standard error 0.9367 0.8426 0.4798 0.9519 0.4881 0.7302 0.9229 0.4731 0.4861
T- value 2.99 2.42 3.46 2.9535 3.2124 2.6911 3.0147 3.2844 3.4673
P- value 0.0174 0.0417 0.0085 0.0183 0.0124 0.0275 0.0167 0.0111 0.0085
Ypc -0.001* -0.0005 -0.0006 -0.001* -0.0002 -0.0007 -0.001* -0.0002 -0.0006
standard error 0.0005 0.0004 0.0003 0.0005 0.0002 0.0004 0.0005 0.0002 0.0003
T- value -2.10 -1.14 -1.72 -2.0658 -0.7942 -1.7512 -2.1337 -0.9219 -1.7629
P- value 0.0688 0.2891 0.1244 0.0727 0.4500 0.1180 0.0654 0.3835 0.1159
HCI -1.44 -1.23 -0.60 -1.4875 -0.7603 -0.5950 -1.3956 -0.6681 -0.6118
standard error 0.7864 0.9017 0.5591 0.8252 0.6191 0.5672 0.7509 0.5643 0.5385
T- value -1.83 -1.37 -1.07 -1.8025 -1.2281 -1.0490 -1.8587 -1.1839 -1.1361
P- value 0.1043 0.2087 0.3149 0.1091 0.2543 0.3248 0.1001 0.2704 0.2888
B20 -0.06* -0.053* -0.029 -0.0604 -0.053*** -0.0391 -0.0603* -0.0535*** -0.0299
standard error 0.0323 0.0265 0.0230 0.0325 0.0158 0.0280 0.0320 0.0158 0.0231
T- value -1.87 -1.99 -1.27 -1.8571 -3.3720 -1.3935 -1.8834 -3.3765 -1.2950
P- value 0.0983 0.0814 0.2413 0.1004 0.0098 0.2010 0.0964 0.0097 0.2314
Un 0.0205 0.0208 -0.0166 0.0206 0.0087 -0.0132 0.0204 0.0069 -0.0154
standard error 0.0517 0.0552 0.0455 0.0522 0.0319 0.0466 0.0513 0.0317 0.0454
T- value 0.40 0.38 -0.36 0.3953 0.2713 -0.2827 0.3980 0.2167 -0.3392
P- value 0.7017 0.7163 0.7252 0.7029 0.7930 0.7846 0.7011 0.8338 0.7432
R Square 0.51 0.39 0.42 0.51 0.2 0.43 0.52 0.21 0.43
227
Appendix IV-F OLS and Empirical Bayes Estimates for URBAN KPK (Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 0.44 0.27 0.97* 0.5333 0.9310 0.5986 0.3530 0.7367 0.9314*
standard error 0.7678 0.8204 0.4443 0.7864 0.5572 0.6187 0.7453 0.4866 0.4451
T- value 0.57 0.33 2.18 0.6781 1.6710 0.9675 0.4736 1.5142 2.0927
P- value 0.5843 0.7476 0.0605 0.5168 0.1333 0.3616 0.6484 0.1684 0.0697
Ypc 0.0002 0.0004 0.00003 0.0002 0.0002 0.0001 0.0002 0.0002 0.0000
standard error 0.0002 0.0003 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002
T- value 0.78 1.24 0.17 0.6795 0.7165 0.6523 0.8659 0.9658 0.2040
P- value 0.4601 0.2494 0.8699 0.5160 0.4941 0.5325 0.4118 0.3624 0.8435
HCI 0.14 -0.05 -0.20 0.0262 -0.3645 0.0301 0.2519 -0.1826 -0.1386
standard error 0.7348 0.8906 0.5127 0.7450 0.6386 0.5372 0.7187 0.5389 0.5079
T- value 0.19 -0.05 -0.39 0.0352 -0.5709 0.0561 0.3506 -0.3387 -0.2729
P- value 0.8504 0.9580 0.7048 0.9728 0.5838 0.9567 0.7350 0.7435 0.7918
B20 -0.038* -0.0402 -0.049*** -0.039* -0.0504** -0.0419** -0.0383* -0.0463** -0.0498***
standard error 0.0186 0.0217 0.0147 0.0189 0.0156 0.0175 0.0183 0.0141 0.0148
T- value -2.09 -1.86 -3.38 -2.1022 -3.2202 -2.3905 -2.0894 -3.2870 -3.3649
P- value 0.0703 0.1007 0.0096 0.0687 0.0122 0.0438 0.0701 0.0111 0.0099
Un 0.0029 0.0164 0.0034 0.0043 0.0099 0.0033 0.0015 0.0067 0.0024
standard error 0.0152 0.0435 0.0144 0.0151 0.0230 0.0139 0.0152 0.0216 0.0143
T- value 0.19 0.38 0.23 0.2867 0.4306 0.2393 0.0971 0.3088 0.1690
P- value 0.8534 0.7159 0.8209 0.7816 0.6781 0.8169 0.9250 0.7654 0.8700
R Square 0.79 0.06 0.77 0.79 0.29 0.79 0.79 0.45 0.77
228
Appendix IV-G OLS and Empirical Bayes Estimates for RURAL BALOCHISTAN (Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.93*** 1.52* 1.68*** 1.9307*** 1.3530** 1.8494*** 1.9332*** 1.3872** 1.6933***
standard error 0.4960 0.6957 0.3745 0.5012 0.4267 0.4695 0.4917 0.4160 0.3770
T- value 3.90 2.19 4.48 3.8520 3.1712 3.9393 3.9314 3.3347 4.4912
P- value 0.0046 0.0604 0.0021 0.0049 0.0132 0.0043 0.0043 0.0103 0.0020
Ypc -0.001** -0.0006 -0.0008*** -0.001** -0.0005* -0.0009** -0.001** -0.0005* -0.0008***
standard error 0.0003 0.0004 0.0002 0.0003 0.0002 0.0003 0.0003 0.0002 0.0002
T- value -3.22 -1.66 -3.41 -3.2069 -1.9421 -3.2074 -3.2224 -2.0486 -3.4121
P- value 0.0123 0.1350 0.0092 0.0125 0.0881 0.0125 0.0122 0.0747 0.0092
HCI 0.019 0.054 0.118 0.0239 0.4018 0.1097 0.0143 0.3491 0.0845
standard error 0.6928 0.8722 0.5902 0.7163 0.5636 0.5412 0.6668 0.5186 0.5682
T- value 0.0273 0.0622 0.2000 0.0333 0.7129 0.2027 0.0215 0.6732 0.1487
P- value 0.9789 0.9519 0.8465 0.9742 0.4962 0.8445 0.9834 0.5198 0.8855
B20 -0.047** -0.047** -0.043** -0.0471** -0.0507*** -0.0470** -0.0470** -0.0496*** -0.0430**
standard error 0.0153 0.0200 0.0136 0.0155 0.0130 0.0141 0.0151 0.0128 0.0134
T- value -3.07 -2.34 -3.18 -3.0337 -3.9043 -3.3339 -3.1050 -3.8779 -3.2180
P- value 0.0154 0.0475 0.0130 0.0162 0.0045 0.0103 0.0146 0.0047 0.0123
Un 0.032 0.038 0.033 0.0323 0.0087 0.0299 0.0326 0.0097 0.0335
standard error 0.0353 0.0502 0.0331 0.0357 0.0293 0.0319 0.0348 0.0291 0.0327
T- value 0.92 0.75 0.98 0.9061 0.2967 0.9363 0.9373 0.3345 1.0271
P- value 0.3841 0.4726 0.3542 0.3913 0.7742 0.3765 0.3760 0.7466 0.3344
R Square 0.72 0.64 0.72 0.72 0.52 0.72 0.72 0.55 0.72
229
Appendix IV-H OLS and Empirical Bayes Estimates for URBAN BALOCHISTAN (Dependent Variable: BNGI) Initial Results HCI-1 HCI-2
Least Squares
Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares Empirical Bayes (HP)
Empirical Bayes (CZ)
Least Squares Empirical Bayes (HP)
Empirical Bayes (CZ)
CONSTANT 1.45** 1.11 1.39*** 1.42** 1.182** 1.373** 1.4803** 1.2309** 1.409***
standard error 0.58 0.73 0.40 0.6124 0.4554 0.5238 0.5592 0.4249 0.3974
T- value 2.49 1.51 3.46 2.3136 2.5943 2.6208 2.6470 2.8969 3.5457
P- value 0.0377 0.1684 0.0086 0.0494 0.0319 0.0306 0.0294 0.0200 0.0076
Ypc -0.0004 -0.0001 -0.0004 -0.0004 -0.0001 -0.0004 -0.0004 -0.0002 -0.0004
standard error 0.0003 0.0004 0.0003 0.0003 0.0002 0.0003 0.0003 0.0002 0.0003
T- value -1.13 -0.24 -1.25 -1.1263 -0.5790 -1.1535 -1.1299 -0.6696 -1.2275
P- value 0.2909 0.8149 0.2472 0.2927 0.5786 0.2820 0.2913 0.5220 0.2545
HCI -0.013 -0.090 0.054 0.0467 -0.0797 0.1046 -0.0593 -0.1049 0.0043
standard error 0.63 0.85 0.50 0.6622 0.5920 0.5099 0.5882 0.5310 0.4824
T- value -0.02 -0.11 0.11 0.0705 -0.1346 0.2051 -0.1009 -0.1976 0.0089
P- value 0.98 0.92 0.92 0.9456 0.8963 0.8426 0.9221 0.8483 0.9931
B20 -0.059*** -0.056** -0.058*** -0.058*** -0.055*** -0.0578*** -0.059*** -0.0563*** -0.0584***
standard error 0.0125 0.0189 0.0121 0.0127 0.0123 0.0122 0.0124 0.0119 0.0120
T- value -4.68 -2.96 -4.80 -4.5896 -4.4847 -4.7380 -4.7694 -4.7135 -4.8577
P- value 0.0016 0.0181 0.0013 0.0018 0.0020 0.0015 0.0014 0.0015 0.0013
Un 0.019 0.031 0.017 0.0179 0.0158 0.0165 0.0201 0.0152 0.0181
standard error 0.0325 0.0489 0.0306 0.0325 0.0276 0.0311 0.0325 0.0276 0.0307
T- value 0.59 0.63 0.56 0.5527 0.5738 0.5304 0.6177 0.5504 0.5878
P- value 0.5737 0.5478 0.5893 0.5956 0.5819 0.6102 0.5539 0.5971 0.5729
R Square 0.84 0.64 0.84 0.84 0.66 0.84 0.84 0.71 0.84