NEW DATA, NEW DOUBTS
REVISED 2013
A Thesis
Presented to the faculty of the Department of Economics
California State University, Sacramento
Submitted in partial satisfaction of
the requirements for the degree of
MASTER OF ARTS
in
Economics
by
Kathleen Lyden
SUMMER
2013
© 2013
Kathleen Lyden
ALL RIGHTS RESERVED
ii
NEW DATA, NEW DOUBTS
REVISED 2013
A Thesis
by
Kathleen Lyden
Approved by:
__________________________________, Committee Chair
Dr Stephen Perez
__________________________________, Second Reader
Dr David Lang
____________________________
Date
iii
Student: Kathleen Lyden
I certify that this student has met the requirements for format contained in the University format
manual, and that this thesis is suitable for shelving in the Library and credit is to be awarded for
the thesis.
__________________________, Graduate Coordinator ___________________
Dr Kristin Kiesel
Date
Department of Economics
iv
Abstract
of
NEW DATA, NEW DOUBTS
REVISED 2013
by
Kathleen Lyden
In 2000, Burnside and Dollar wrote a seminal paper declaring their finding that
international economic aid works to increase economic growth in less developed
countries where there is good policy. However, in 2003, Easterly, Levine and Roodman,
disputed this result, finding the inclusion of additional countries into the sample and a
broadening of years covered by the study diminish the robustness of the policy variables’
significance. This thesis will recreate the work of Easterly, Levine and Roodman, and update it where they left off in 1997 with the most current data to determine if these results
still hold true. Results suggest that policy is in deed significant to growth, irrespective of
aid, though the aid-growth relationship is ambiguous, as is any interaction aid and policy
have with growth.
_______________________, Committee Chair
Dr Stephen Perez
_______________________
Date
v
TABLE OF CONTENTS
Page
List of Tables ......................................................................................................................... vii
List of Figures ....................................................................................................................... viii
Chapter
1. INTRODUCTION ………………...…………………………………………………….. 1
2. LITERATURE REVIEW ................................................................................................... 4
Economic Growth Theory............................................................................................ 4
Economic Aid in Economic Growth Theory ............................................................... 8
Policy in the Aid-Growth Theory Dynamic............................................................... 14
3. DATA .............................................................................................................................. 18
4. EMPIRICAL METHODOLOGY .................................................................................... 40
5. ANALYSIS ...................................................................................................................... 45
6. RESULTS, ROBUSTNESS TESTS, AND INTERPRETATIONS .................................. 53
Final Estimation Details............................................................................................. 53
Evaluations of Robustness ......................................................................................... 56
7. FINAL CONCLUSION .................................................................................................... 68
Appendix A. Descriptive Statistics ...................................................................................... 72
Appendix B. Initial Result Tables ........................................................................................ 73
Appendix C. Sign Table ....................................................................................................... 81
Appendix D. Final Result Table .......................................................................................... 82
Bibliography ........................................................................................................................... 83
vi
LIST OF TABLES
Tables
Page
1.
Table of Variables ………………………………… .………………………………. 18
2.
Countries included in the Sample………………….… . ……………………………. 19
3.
Time Periods in the Sample……………………….…………………………………. 20
vii
LIST OF FIGURES
Figures
1.
Page
Growth-Aid*Policy Plot ……………………………….………………………………. 58
viii
1
Chapter 1
INTRODUCTION
As the history of the world has progressed, each stage of human interaction has
brought with it a new kind of development strategy and concern. Yet, the method of
donors allocating funds in such a consistent and seemingly benevolent fashion is one that
had not existed until the end of World War II. Prior to that, aid programs, the patterns of
one country contributing to the financial wellbeing of another, primarily consisted of
conquest, imperialism and colonization. After World War II, the primary concerns were
the rebuilding and economic stability of the nations devastated by the War and the
transition of the former colonies into independent nation states. With the ascendance of
embedded liberalism and Keynesian economics, the aid and development efforts focused
on the construction of national institutions to support capitalism and open markets in the
future.
In this time period, the concept of poverty was also born from the World Bank and
Robert McNamara’s call for the alleviation of poverty around the world (Birdsall &
Londono, 1997). Thus, from this point on, a large portion of bilateral and multilateral aid
has been focused on improving the lives of the poor around the world, and developing
their national economies to raise their standards of living. Yet the image of the poor was
that of helpless, unemployable individuals, too lazy or ignorant to help themselves (Hiatt
& Woodworth, 2006). Thus, governments were seen as the only appropriate mechanism
to receive and implement aid programs for the destitute and economic growth and
2
development typically are seen as the tools in which to bring about the desired change.
The idea behind allocations at this level was that the benefits of the development of
national economic infrastructures would trickle down to all the individuals in the society.
However, this has not tended to be the result of such aid. Large-scale infrastructural
projects are rarely of benefit to or are seen by the poor they are supposed to be helping
(Maren, 1997), and tend to have little real impact on the development and growth of the
domestic economy (Yunus, 2003). Finally, the policies and institutions seen as necessary
for good governance are largely absent from these impoverished economies. For the
developing world, this means the inefficient allocation of resources, overshadowed by
powerful domestic interest groups, weak governments and corruption (Pietrobelli &
Scarpa, 1992).
As a result, much of the conversation surrounding aid has been about how to make it
more efficient, how to transform the methods and modes of aid allocation so that it can
better achieve its goals. In doing so, one must have a clearer understanding of what
causes aid to not have the intended effect. Understanding the basic economic growth
model, and how aid fits into that model, provides a crucial juncture for understanding
these relationships. In 2000, Craig Burnside and David Dollar brought forth one of the
seminal works on the role of aid in economic development. They found evidence that
supported the hypothesis that aid better supports economic growth in countries with good
policies. In 2003, William Easterly, Ross Levine and David Roodman repeated the work
of Burnside and Dollar, adding to and revising the data. Their results failed to show the
same robustness for good policy’s influence on aid’s impact of economic growth.
3
Both these studies cover 1970 through some portion of the 1990’s. The purpose of
this research will be to expand the data to include all that is currently available and retest
the hypothesis that good policy positively influences aids’ impact on economic growth.
As economic growth and development often mean increased living standards, increased
political stability, increasing numbers of markets to trade to and import from, especially
for poorer economies, good policies are a way national governments can signal to the
outside world—such as donors, intergovernmental organizations, nongovernmental
organizations, businesses, entrepreneurs, tourists, immigrants—that it would be safe and
well advised to spend their money in-country.
The rest of this section will discuss how this research will proceed. The next chapter
of this thesis contains a literature review of the economic growth theory, the role of
developmental aid in growth theory, and how policy has entered into the aid-growth
dynamic. Following that, Chapter 3, about the data, will describe the variables, where
they originated, and how they are used in the analysis. The fourth chapter will be a
discussion of the methodology used to analyze the data. After that will be the initial
analysis in Chapter 5 and further analysis with robustness tests and interpretations in
Chapter 6, followed by the final conclusion in Chapter 7.
4
Chapter 2
LITERATURE REVIEW
Economic Growth Theory
The basic form of economic growth theory calls for growth to be derived from the
relationship of output to capital and labor inputs such that some relationship between
increasing some portion of inputs will result in the increased output. The Neoclassical
models of the 1950s and 1960s (Solow, 1956; Swan, 1956) call for a specification of the
economy as closed, without technology, and with unitary elasticity of inputs which have
a flexible relationship. The savings rate, which is constant and exogenous, influences the
capital-output ratio. Differences in the capital-labor ratio lead to different growth paths
over the long-run. Where capital-labor ratios are higher, long-run output per person tends
to be higher; and where they are lower, long-run output per person tends to be lower.
Although, within each growth path, population growth - and thus the labor force growth
rate, which is also exogenously determined - is the driver to economic growth.
However, these models were not seen to accurately explain the rapid growth of the
more developed economies over time. Thus, scholars sought to incorporate some
quantification for technology that would provide this additional explanatory power to the
model. Ramsey (1928) developed a theory of household utility maximization that
established the ratio of savings to consumption and labor to leisure time that a household
chose to have. Cass (1965) and Koopmans (1965) applied this constraint to the SolowSwan theory of growth to better explain saving and consumption decisions, capital
5
accumulation, and through those the growth of the economy. Learning-by-doing was
added (Arrow, 1962; Sheshinski, 1967) to the Neoclassical growth model to represent the
accumulation of knowledge as a measure of the level of technology that could not be
contributed to increases in the capital-labor ratio. Sheshinski (1967) found that the higher
the level of investment in the accumulation of knowledge, the greater was the speed of
technical progress and production efficiency. Thus, increasing output returns to
investment were possible. Nelson and Phelps (1966) reach a similar conclusion by
treating technology as an element that augments labor through education. Increases in
education units per worker lead to higher levels of productivity and a greater propensity
for those workers to encourage the dissemination of more technology, increasing the
speed with which new ideas spread. Likewise, Mankiw, Romer and Weil (1992) find that
technology augmented labor, via schooling levels, does a good job of explaining real
world growth paths and convergence. It has further been suggested that, while education
levels unequivocally contribute to the increased efficiency of workers, and through that
generate higher levels of output and increase income levels, the education of female
students provide greater returns than the education of male students (Shultz, 1993).
However, these returns may often be over-looked as not productive to the labor market.
On the other hand, Bils and Klenow (2000) find that there is not sufficient causal
evidence to explain the level of growth attributed to knowledge accumulation through
education levels. Rather, they suggest that it is likely that economic growth contributes
significantly to education levels, as well as the likelihood that some other omitted
6
variable explains the linkage between initial capital stocks in 1960 and economic growth
rates from 1960 to 1990.
When compared side by side and with the theories that support technological progress
through learning-by-doing, it has been found that models emphasizing human capital
accumulation through schooling do not allow for convergence and do not explain
historical growth paths as well as the models that explain human capital accumulation
through learning-by-doing (Lucas, 1988). Since different kinds of production have
different potentials, the variance in the accumulation of human capital specialized in
certain industries allows for the possibilities of substantial differences in the economic
growth paths that human capital accumulation through education cannot explain. No
matter the method of the accumulation, several scholars have found that human capital
stocks play a significant role in the ability of an economy to experience higher levels of
long-run growth (Uzawa, 1965; Mulligan & Sala-i-Martin, 1993; Cabelle & Santos,
1993; Benhabib & Spiegel, 1994; Romer, 1990).
With these conversations about technology and the investment necessary to
encourage it, the drivers in economic growth theory shifted from being thought of as
exogenous, or outside of the model, to being endogenous. With respect to technology, the
intentional investment in human and physical capital, as well as the accumulated stocks
of both kinds from all prior time periods, mean that the level of technology and the
influence it has on the level of output, in large part, rely in decisions and outcomes in
current and prior time periods (Romer, 1986). This can be further augmented through
specialization of an industry or economy and the opening of the economy to international
7
trade (Romer, 1987, 1990). Krugman (1979) introduces the need for continual innovation
alongside economic growth in order to ensure maintenance and enhancement of living
standards in the long run. Further research supports the theory that innovation and
imitation endogenously increase output growth (Segerstrom, 1991; Aghion & Howitt,
1992). However, they also suggest that innovation and imitation can have a negative
impact on future levels of innovation and imitation due to large relative costs, the
inability to fully capture the returns of innovation, and the prospect of negating rents on
current innovations when future innovations become more marketable. Jones (1995) finds
that this process of innovation is endogenous in the short run, relying primarily on the
level of skilled labor. However, in the long run, the expansion of this skilled labor force
is dependent on the exogenous population growth to increase the size of the labor force as
a whole.
An alternate explanation for the sustained high levels of growth that have been
observed is that economic growth is linked to political development of democratic
institutions. Friedman (1962) cites that the historical trend for the vast majority of human
history was one of servitude and poverty. However, with the rise of economic
opportunities associated with the free market and capitalist institutions, political liberties
flourished. As households were permitted to interact in cooperative transactions to the
mutual benefit of both parties, political rights developed, and through the two, prosperity
increased. Huber, Rueschemeyer and Stephens (1993) suggest that it is not just the rise in
living standards that has led to greater political freedoms, but rather the economic and
8
political environment that alters the social and class structure of a society through
industrialization and urbanization that enable this dual development.
Economic Aid in Economic Growth Theory
“[Economic] growth theory became excessively technical and steadily lost contact with
empirical applications. In contrast, development economists, who are required to give
advice to sick countries, retained an applied perspective and tended to use models that
were technically unsophisticated but empirically usefully. The fields of economic
development and economic growth drifted apart, and the two areas became almost
completely separated,” Barro & Sala-i-Martin, 1995; p12.
Aid, itself, often gets a bad wrap. With solicitations from nongovernmental
organizations on how you can save a child for less than the price of coffee every day,
media reports about food aid being diverted or rotting before distribution, and
investigations into improprieties within and high operations costs at private aid groups,
sentiment towards aid organizations is running low (Maren, 1997). This reputation is not
helped by the facts that aid does not often align with recipient need, accountability is
typically low, and maintenance falls to the recipient who cannot afford to maintain and
whose requests for maintenance funds fall on deaf ears (Easterly, 2006; Wallace et al,
2007; Chandra, 2008). Further, the volatility of aid tends to increase uncertainty
throughout an already delicate economy (Lensink & Morrissey, 2000). Additional
conditionalities and economic restructuring tools such as trade liberalization and
privatization have their share of problems too. Trade liberalization is supposed to put all
9
trading partners, rich and poor, on a more level playing field, though that rarely occurs.
The requirement frequently hinders the poorer country from maintaining budget revenues
that are often reliant on tariffs for half their income, reduces business activity, and
increases unemployment (Chang, 2008). The idea behind privatization is that private
firms operate businesses more efficiently than governments do. However, state ownership
in poorer countries often occurs because no other enterprise has the resources or
knowledge to better meet the public need. Privatizing, only a change of ownership, not
only fails to address the issues that it intends to correct, but often generates deeper
problems (Chang, 2008). Structural adjustment programs as a whole have not had the
intended results and may lead to increased instability (Burnell, 1997; Cohen & Easterly,
2009; Lensink & Morrissey, 2000). However, the number of people living in absolute
poverty and the percent of the population living in absolute poverty are lower now than
they were in 19901. The world life expectancy at birth has risen for both males and
females2. The percent of the population with access to improved sanitation facilities and
the percent in rural areas with improved access to a water source have risen3. Infant
mortality per 1,000 live births has declined4. Therefore, something must be going right.
Aid as a means of encouraging economic growth and development first became an
arena of interest following the Second World War and the extension of economic
assistance, through grants and loans, to countries that were devastated by that war. With
World Bank. “Poverty & Equity Data: World Development Indicators, 2013” Accessed July 2013 <
http://povertydata.worldbank.org/poverty>
Global Poverty Indicators – People living on less than $1.25/day (PPP) in millions: 1,908 (1990); 1,215 (2010)
Poverty headcount ratio at $1.25/day (PPP) as % of population: 43.1% (1990); 20.6% (2010)
2 Females: 70 (2003); 72 (2011); Males: 65.8 (2003); 67.9 (2011)
3 Improved sanitation facilities: 58.2% (2003); 62.6% (2011); improved water source, rural: 74.3% (2003); 80.4% (2010)
4 Mortality rate, infant: 46.8% (2003); 36.9% (2011)
1
10
the creation of the World Bank and the International Monetary Fund, the world had
multilateral organizations that could help promote such rebuilding. Then, after the
majority of the reconstruction was under way, these institutions turned their attention
towards the newly independent or soon to be independent countries constructed with the
decline of colonialization. As a means of maintaining historical markets for colonial
masters, as well as encouraging the expansion of natural resource suppliers and final
goods markets, newly independent countries represented a bonanza for developed
economies (Chenery & Strout, 1966). Further, with the re-innovation of the World Bank
in the McNamara era, foreign aid became a vehicle for alleviating poverty and raising
living standards for the poorest of the poor, for helping to prevent the spread of
communism through the expansion of democracy, which required mobilization of society
that only elevation of living standards could instigate.
Within the growth framework, the role of aid remains ambiguous. Some scholars find
that aid is positively related to growth while others find it is either never significant or is
negatively related to growth. Much recent literature expresses conditionalities of aid,
which, when present, lead to a positive aid-growth relationship, or find that aid indirectly
impacts growth through other variables that directly influence growth, such as investment
or government expenditure.
Early aid-growth theory centers on the idea that aid is used to fill gaps that occur
within less developed economies. By then filling those gaps, aid translates to economic
growth. The first gap that aid addresses is the gap most poor economies face between
savings and investment. The savings rate in less developed countries tends to be lower, ad
11
thus is not sufficient to meet the investment needs or desires. Here, aid supplements
savings and a higher level of investment can occur than would naturally be the case. The
second gap occurs between imports and exports. In this instance, the income generated by
exports, typically lower level manufactured and primary goods, does not cover the
expenses for the desired imports. Aid can further enable the opportunity for growth by
filling this gap. Using the two-gap model based on the Harrod-Domer growth model,
Chenery & Strout (1966) found that aid positively impacts growth through increased
savings and investment, higher levels of human and physical capital, and institutional
development and growth. Adelman and Chenery (1966) found similar results in their case
study for Greece. Papanek (1972, 1973), also using the two-gap model, found aid
explained more growth than any of the other variables. Fayissa and El-Kaissy (1999)
found that aid to be positive and significantly related to growth, particularly when human
capital is considered.
Griffin and Enos (1970), using a one-gap savings model and testing only aid, found
that aid does not positively affect growth, but may divert domestic savings and
investment with it. These findings were challenged due to model and variable
misspecification (Snyder, 1990). However, Bornschier et al (1978) found that in the short
run aid increases growth, but in the long run aid has a negative impact on growth. Mosley
(1980) used a similar model and found that aid is not significant for the whole sample of
less developed countries, but it is for the poorest. Further, the aid-savings relationship
was negative. This, however, is likely related to the fact that poor countries have trouble
saving, and has nothing to do with aid. Singh (1985) found that aid is not significant
12
when government intervention is considered. Islam (1992) found that aid as a whole was
not signifiacnt to growth, but that loans are better than grants and lagged one time period
food aid is better than commodity and project aid. In Cameroon, Mbaku (1993) found
that neither aid as a whole nor aid separated into grants and loans has any significant
impact on growth, though domestic savings does. Countering this, Giles (1994) found
that when aid variables are lagged one time period, grant aid and loan aid are both
positive and significant in contributing to growth.
Since the models based primarily on savings or investment and aid, much of the
research that has followed has been a hodge podge of models and variables. Some models
will include savings or investment variables while others will not. Some will include
capital formulation, labor force or population variables. Regime type, economic and
political risk, and education levels have been other variables considered, as well as
assassinations, ethnolinguistic fractionalization, regional or colonial variables, and
climate related variables. This has led to a lack of cohesion and consensus on what is the
real relationship between economic aid and growth. Durbarry et al (1998), Lensink and
White (2001), and Radelet et al (2004) found that aid is positively related to growth, but
with diminishing returns. Gounder (2001) found that aid and all its components are
positive and significant in contributing to growth in Fiji. Gomanee et al (2002) found that
while aid is not immediately effective, lagged results show that a 1% increase in aid
translates to a 0.33% increase in growth through investment and government
consumption transmission mechanisms. Addison et al (2005) found that aid works
broadly to increase growth that the criticism, that aid is harmful has not been supported
13
by the research, and that aid effectiveness is not the reason for lack of growth in SubSaharan Africa. Minoiu and Reddy (2010) found that, when aid is separated between
developmental and non-developmental aid, developmental aid has strong long-run
benefits to growth, such that even a small amount of the right kind of aid can have longlasting effects.
Several studies have found conflicting results. Boone (1996) found that aid does not
increase growth or investment, and that it does not benefit human development
indicators. Yet, it does increase consumption and the size of government. Knack (2000)
found that increased aid decreases government quality, which could help explain how aid
might negatively impact growth. Islam (2003) found that aid has a negative relationship
with growth, except when the regime is totalitarian. With totalitarian regimes, aid has a
positive relationship to growth with diminishing returns. Rajan and Subramanian (2005)
found that aid reduces competitiveness as expressed through more labor-intense, exportoriented manufactuing. As a result, labor, wages and growth decline. Friedman (1958)
found economic aid to be only a political tool, and that capital given as aid is often
wasted, contributing nothing to growth. Poor protection of property rights and the
maintenance of international trade protections are the greatest barriers to economic
prosperity and growth. Chenery et al (1979) found that aid as a whole has failed to bring
growth, though actual country-to-country results vary. Chatterjee et al (2007) found that
aid had no effect on growth due to fungability.
14
Policy in the Aid-Growth Theory Dynamic
In the last decade of the twentieth century, scholars and policy makers were looking
at growth, aid and the poverty statistics they had at their disposal, trying to find a way to
make better tools for aid efficiency and measurement of real outcomes. There had been
an increasing move in the literature to consider corruption, the quality of bureaucracy, the
business environment and other economic and political issues that might affect growth.
Burnside and Dollar (1997) published the first well-documented findings that policy,
defined as an index of a balanced budget variable, inflation, and a variable for economic
openness, is positive and statistically significant in determining growth. The idea behind
the index is that the budget variable instruments for fiscal policy, inflation instruments for
monetary policy, and economic openness instruments for trade policy. Taken together in
a combined index, they will relay capacity of the economy to absorb, allocate, and utilize
aid funds so that they benefit growth. These findings led to the World Bank’s Assessing
Aid: What Works, What Doesn’t, and Why (1998) report and corroboration that aid will
not always lead to economic growth, but is dependent on ‘good’ economic policies as
outlined by Burnside and Dollar. The report also suggested that growth is dependent on
political institutions, such as the rule of law, the level of corruption, the function of the
bureaucracy and accountability in government. As a result, the Bank formulated policy
recommendations to donor governments and agencies that economic aid be given to
recipients with good policy, and that recipients with poor policies be offered aid only as a
reward after changing their policies. In 2000, Burnside and Dollar published their seminal
15
paper further substantiating findings that policy has a role in the economic growth model.
They clarified that, while they found that aid has a positive and significant impact on
growth when good economic policies are in place, it has no significant impact when poor
policies are present. They further suggest that it is really the good policies that are most
important for growth, and that aid can only be most effective when allocated to recipients
with good policies. Additional research has found that the allocation of aid is positively
related to growth when policies and institutions of good governance are stronger (Ali
&Isse, 2005; Dollar & Levin, 2005; Dollar & Levin, 2006). Collier and Dollar (2001,
2002) further contribute to this with findings that suggest that policy reforms and
efficient aid allocations to recipients that continue to make these good policy economic
reforms contribute to growth such that poverty could be halved by 2015. Additional
research by Cogneau and Naudet (2007) support the findings that good policy is
necessary to allow aid to be efficiently allocated within the recipient economy for higher
levels of growth and poverty reduction. This relationship still holds when negative shocks
are introduced (Collier & Dehn, 2001) and in post-conflict societies (Collier & Hoeffler,
2002), and the timing of certain policies and aid allocations seem to become more
important. Additional results suggest that policy helps aid be more effective, but that it is
not a conditionality (Dalgaard & Hansen, 2001). Hudson and Mosley (2001) found
somewhat conflicting evidence that good policy matters in enhancing growth, but that it
does not necessarily make aid any more or less efficient.
Other scholars have criticized the World Bank findings and policy suggestions, and
additional research suggests that good governance policies and institutions do not
16
necessarily improve the relationship between aid flows and economic growth. Economic
aid has been shown to increase growth and reduce poverty, while these economic good
governance policies have no theoretical link to growth (Degnbol-Martinussen &
Engberg-Pedersen, 2003). Furthermore, allocations should be made based on need and
the ablility to align donor and recipient needs, not with some ideal of past performance
(Degnbol-Martinussen & Engberg-Pedersen, 2003; Hermes & Lensink, 2001).
Additionally, punishing poor performers for past behavior is both unfair and possibly
presumptuous (Degnbol-Martinussen & Engberg-Pedersen, 2003; Hermes & Lensink,
2001). Poor governments rarely have the resources to control very many aspects of their
government and economy. However, when they do, donor governments and policy
analysts will not necessarily have a better understanding of the local environment to help
create and implement workable plans that will better meet the needs of the economy and
the donors’ expectations. There is evidence that good governance policies and institutions
do not necessarily improve the relationship between aid flows and economic growth,
more specifically that the aid-growth relationship is positive and significant irrespective
of policy and that policy does not add anything to the relationship (Hansen & Tarp, 2000,
2001; Guillaumont & Chauvet, 2001; Morrissey, 2001; Easterly, Levine, & Roodman,
2003; Ram, 2004; Rajan & Subramanian, 2008). Further findings suggest that not only is
policy not relevant in the relationship, but that a geographical or climate variable does
contribute significantly to determining aid on growth (Dalgaard, Hansen & Tarp, 2004;
Roodman, 2004).
17
Further, when considering indicators of good governance policies, the literature
suggests donor inconsistency. First, it is typically accepted that the quality of governance
institutions are a determining factor in the level of corruption in an economy, as low
quality governance tends to be indicative of weak governments. Further, weak
governments tend to have little control over their agencies and departments, (Shleifer &
Vishny, 1993). However, when donors allocate aid, they frequently do not reward corrupt
recipients much differently than the non-corrupt, (Alesina &Weder, 2002). While there is
evidence that this varies among donors, when there is a strong push for good governance,
accountability, and transparency in the rhetoric of aid allocations, the failure to
differentiate between actual quality of recipient governance sends the signal that the
donors are not really serious about the quality of governance or the level of corruption.
Also with respect to the donor motives and considerations in aid allocations, evidence
shows that the majority of the time, the self-interest of donors outweighs the needs of the
recipients and the stated policy requirements and preferences, such as high quality
governance and low levels of corruption, of the donors, (Hook, 1995; Alesina & Dollar,
2000; Zanger, 2000; Neumayer, 2003).
18
Chapter 3
DATA
The variables used in this research are those that were used in both the Burnside and
Dollar (2000) and the Easterly, Levine, and Roodman (2003) papers, though there are
slight deviations between those two works, and between those works and this one. All of
the original data from both studies are available and the results of Easterly, Levine, and
Roodman have been duplicated with files available on the Center for Global
Development website.
Table 1: Table of Variables
Expected Sign
(+)
(+)
(+)
(-)
(-)
(+)
(-)
(-)
(-)
(+)
(+)
( +/ - )
(-)
( +/ - )
Instruments
to Aid/ aid*policy
(+)
(-)
(+)
(+)
(+)
Name
GDPG
AID
POLICY#_b1
BB
INFL
SACW
GDP
GDP 1970
SSA
EAsia
ICRGE
M2 (-1)
ETHNF
_Period*
DN1900
Outliers#_b1
Variable
Per Capita GDP Growth
Net total ODA, constant 2011 US millions, from all donors
Policy
Budget Surplus as % of GDP
Inflation - Consumer Price Index (bd or elr)
Inflation - GDP deflator (kl)
Total imports plus total exports over per capita GDP, in constant 2005 US$
Initial GDP per Capita for each time period in 2000 US$
Initial GDP per Capital in 1970 in 2000 US$
Sub-Saharan Africa Dummy variable
East Asia Dummy variable
Institutional Quality
Lagged money supply
Ethnic Fractionalization
Time Period dummies
Dummy for low income countries
Aid*Policy falling>±2SD from mean dummy variable
Primary Source
Range of available years
World Bank WDI
1961-2011
OECD.Stat DAC2a
1960-2011
Constructed from Data
1966-2009
ELR; IMF World Econ Outlook
1960-1997
World Bank WDI
1960-2011
World Bank WDI
1960 - 2011
Penn World Tables
1960 - 2010
World Bank WDI
1960-2011
World Bank WDI
1970
ELR
Constant
ELR
Constant
ELR
Constant
World Bank WDI
1960-2011
ELR
Constant
Constructed from Data
Constant
ELR
Constant
Constructed from Data
Constant
LPOP
ARMS-1
FRZ
EGYPT
CENTAM
Natural Logarithm of Population
Lagged Arms imports as % of total imports
Dummy for Former French Colonies
Dummy for Egypt
Dummy for Central American countries
World Bank WDI
World Bank WDI
ELR
ELR
ELR
1960-2011
1960-2011
Constant
Constant
Constant
19
Countries in the Sample
There are 63 countries in this sample, 56 from the original Burnside-Dollar (BD) data
set. Easterly, Levine, and Roodman (ELR) added the other seven. There was one
additional country, Somalia, in the sample, but it was dropped for lack of consistent data.
For the first set of estimations, only the BD countries are used. However, in all of the
other estimations, it was the full sample of countries.
Table 2
Countries included in the Sample
Ghana
*
Algeria
Guatemala
Argentina
Guyana
Paraguay
Bolivia
Haiti
Peru
Botswana
Honduras
Philippines
Brazil
India
Senegal
Burkina Faso
Indonesia
Sierra Leone
Cameroon
*
Chile
Colombia
*
Iran
*
Papua New Guinea
South Africa
Jamaica
Sri Lanka
*
Jordan
Syria
Congo, Democratic Republic of the
Kenya
Tanzania
Congo, Republic of the
Korea, Republic of
Thailand
Costa Rica
Madagascar
Togo
Cote d'Ivore
Malawi
Trinidad & Tobago
Dominican Republic
Malaysia
Tunisia
Ecuador
Mali
Turkey
Egypt
Mexico
El Salvador
Morocco
Uruguay
Ethiopia
Myanmar
Venezuela
Gabon
Nicaragua
Zambia
Gambia
Niger
Zimbabwe
Nigeria
*
*
Countries Not in original BD study
Pakistan
*
Uganda
20
Time periods in the Sample
This data set covers the years from 1962 – 2009, with annual data converted into
four-year averages. The original BD research covered time periods two through seven
(1970-1993), as shown in Table 3. This set of time periods was tested twice, once with
only the BD sample countries and then with the full sample of countries. Next, the ELR
research covers time periods two through eight (1970-1997). This set of time periods was
only tested with the full sample of countries. Finally, the extension of the data this
research offers includes time periods one through eleven (1966-2009). Time period zero
(0) was dropped for lack of sufficient data. This set of time periods was also only tested
with the full sample of countries.
Table 3: Time Periods in the Sample
Period
0
1
2
3
4
5
6
7
8
9
10
11
Period start
1962
1966
1970
1974
1978
1982
1986
1990
1994
1998
2002
2006
Period end
1965
1969
1973
1977
1981
1985
1989
1993
1997
2001
2005
2009
Economic Growth
First, the dependent variable tested by all three works, GDP growth, is the annual per
capita growth of gross domestic product (GDP). For the ELR paper, this data came from
21
the World Bank World Development Indicators database, 2002. For this work, the
majority of this data came from the World Bank World Development Indicators
Database, 2013. Missing data was also calculated from per capita GDP values5 or
extracted from the ELR database6. As compared to the data in the original studies, the
economic growth data in this study has a 0.9583 correlation value with the economic data
in the BD study and a 0.9840 correlation value with the ELR data.
The mean annual economic growth is 1.66%, with the median of 1.82%, indicating
slightly higher positive growth and fewer instances of low and negative growth across the
sample. The standard deviation of economic growth values is 3.36, indicating that 95% of
all economic growth values for this sample should fall within a range ±6.72 above or
below the mean, or between -5.06% and 8.38%. The minimum growth for this sample
was -13.21%, while the maximum was 21.69%.
Developmental Aid
The aid variable used in the original works was Effective Development Assistance
(EDA) as a percent of GDP per capita. EDA is a measure of international development
assistance introduced by Chang, Fernández-Arias and Servén in “Measuring Aid Flows:
A New Approach,” (1998). This value for aid is unique because it estimates a more
accurate grant-value to official loans by discounting them for the anticipated future
interest payments, adding this to the total grant value, while excluding technical
Countries for which data was retrieved this way are: Myanmar 2005-2010; Niger 1961-1965; Sierra Leone
1961-1965
6 Haiti 1970-1989; Tanzania 1982-85
5
22
assistance, which they contend really does more to benefit the donor than the recipient.
The full database of results from this work, for 1975 – 1995, can be found on the World
Bank website. Easterly, Levine and Roodman found that there was a high level of
correlation between the EDA and Net Official Development Assistance (Net ODA).
Thus, they used Net ODA to extrapolate EDA for 1970-1974 and 1996-1997 to complete
the years of aid data necessary for their sample. When calculated from the original
Chang, Fernández-Arias and Servén data and current Net ODA data, obtained from the
Organization for Economic Co-operation and Development (OECD), the correlation
between the two values for all countries in this sample for 1975 – 1995 is 0.8855. It is not
clear whether EDA values are in current or constant million US$. However, when
adjusting Net ODA values to current year rather than 2011 millions of US$, the
correlation between the two values increases to 0.95627. Thus, the decision was made to
use Net ODA as the measure for aid in this research, rather than try to extrapolate for
over 20 years worth of data or reconstruct all the appropriate loan detail information that
would enable the extension of the EDA database. These values were then adjusted to
2000 US$ and divided by the 2000 US$ GDP value obtained from the World Bank’s
World Development Indicators, 2013, database to get aid/GDP. The aid data in this study
has a correlation value of 0.8113 with the BD aid variable, and 0.8188 with the ELR aid
variable.
The average country in an average year received $638.5 million in net official
development assistance. With the median net ODA of $359.8 million, this suggests most
7
Bureau of Labor Statistics CPI inflation calculator was used in adjusting all current year to constant year calculations
23
countries received much lower sums of net ODA while a few received much larger. The
standard deviation of $883.4 million indicates that 95% of the sample country aid receipts
should fall within the range of ±$1766.8 million of the mean, or between -$1150 million
and $2405.3 million. The lowest sample observation is -$392.3 million.
In terms of percent of GDP, the mean net ODA is 3.29%. This suggests that in an
average year, the average recipient country could expect net ODA equivalent of about
3.29% of its GDP. However, with a median of 1.67%, the results indicate that it is likely
that most countries receive much smaller aid allocations in terms of their GDP, while a
few others receive much larger portions. The standard deviation of 4.27% and minimum
Net ODA observation of -0.07% support that. Given the standard deviation, 95% of the
countries’ Net ODA receipts should fall between ±8.54% GDP from the mean, or
between -5.35% GDP and 11.83% GDP.
Budget Surplus/Balanced Budget Variable
There are three variables that go into the creation of the policy index variable, with
the federal government budget variable being the first. This variable is expressed as total
revenue minus total expenditure as a percent of GDP. Thus, a negative value indicates a
budget deficit, while a positive value indicates a budget surplus. Data for ELR was
retrieved from the World Bank World Development Indicator database, 2002, with
additional data extrapolated from the IMF International Financial Statistics database,
2002. For this research, the data for the earlier years of this variable was taken from the
ELR database, while data for later years came from the IMF World Economic Outlook
24
Database, 20138. The data in this research has a correlation value with the BD data of
0.9213, and with the ELR data of 0.9989.
The mean budget variable is -0.03, indicating that the average government, in an
average year, will have a 3% budget deficit. The standard deviation for this variable is
0.05, meaning that 95% of all budget values should fall within ±0.10 of the mean, or
within the range of -13% and 7%.
Inflation
The second variable that goes into the policy index variable is inflation. ELR define
this as the natural logarithm of one plus the inflation variable. Both BD and ELR have
selected to use the consumer price index (CPI) as their measure of inflation. ELR obtain
this data from the World Bank World Development Indicator database, 2002. In this
research, both the GDP deflator and the CPI are used as measures of inflation. The GDP
deflator was added for several reasons. First, it is more widely available than the CPI.
Second, as a measure of GDP inflation, rather than consumer prices, it seems a more
reasonable measure to associate with GDP growth as the GDP deflator addresses inflation
within the economy as a whole, while the CPI only looks at consumer-related inflation
(Bureau of Labor Statistics, 2011; Politonomist, 2009). The CPI measures the prices of
given set of durable and non-durable consumer goods like food and housing, while the
GDP deflator considers all of the components of the economy that make up the GDP
(Bureau of Labor Statistics, 2011; Politonomist, 2009; United Nations, 2009). Thus, the
Additional data from PRS Group Annual Country Risk Reports: Egypt 1998-99; Gambia 1998-99; Guatemala 1998-99; Honduras 199899; Mali 1998-99; Nicaragua 1998-99; Nigeria 1998-99; Peru 1998-99; Sierra Leone 1998-99; South Africa; 1998-99; Trinidad & Tobago
1998; Uruguay 1998-99; Zambia 1998-99
8
25
GDP deflator includes inflationary influences from not only consumption, but
investments, net exports and government expenditures as well (Bureau of Labor
Statistics, 2011; Politonomist, 2009).
Both CPI and GDP deflator data for this research come from the World Bank World
Development Indicator database, 2013. Missing data was retrieved from the ELR
database9. When correlated with the original BD and ELR inflation variables, the CPI
inflation variable in this research had correlation values of 0.8628 and 0.8991,
respectively. When the GDP deflator variable was correlated with the original BD and
ELR inflation variables, the correlations values were 0.9738 and 0.7254, respectively.
The average CPI inflation value is 44.12%, with a median of 9.19% and a standard
deviation of 304.72%. So the range within which 95% of all sample values should fall is
between -565.32% and 653.56%. The minimum value in the sample is -3.62%, while the
maximum is 6251.5%. For the purposes of the data as used in this analysis, the natural
logarithm of (1+CPI), the mean value is 0.174, with a median of 0.088 and a standard
deviation of 0.364. Thus, 95% of the sample should fall within ±0.728 of the mean, or
between -0.554 and 0.902. The sample minimum is -0.037 and the maximum is 4.151.
When looking at the GDP deflator inflation variable data, the pattern of relatively
high mean, relatively low median, and large standard deviation are similar. The GDP
deflator mean is 53.85%, with a median of 8.89% and a standard deviation of 378.89%.
The range within which 95% of the sample should fall is between -703.93% and
811.63%, while the sample has a minimum of -10.06% and a maximum of 7014.81%.
Observation (country and time period) from this source were: Chile 1-8, Costa Rica 1-5, Nicaragua 2-8; Sierra Leone 1-8; Uganda 4;
Venezuela 1-8.
9
26
For the purposes of the data as used in this analysis, the natural logarithm of (1+GDP
deflator), the mean value is 0.178, with a median of 0.085 and a standard deviation of
0.406. Thus, 95% of the sample should fall within ±0.812 of the mean, or between
-0.634 and 0.990. The sample minimum is -0.106 and the maximum is 4.265.
Openness/Sachs-Wagner Variable
The final variable that goes into the policy index variable is economic openness. BD
use the Sachs-Wagner variable, which rates an economy as open or closed based on the
black market exchange rate premium, the levels of tariff and non-tariff barriers, whether
the government is socialist, and the level of government involvement in the primary crop
export market. ELR use the same measure for openness, though up-date it to more
accurately reflect the true level of economic openness. The trade information was
obtained from UNCTAD, the black market exchange rate premium from the Global
Development Network database and assorted other resources, the export market
information from the IMF, and the country’s status as Socialist from the CIA. As several
of these variables are not available in current or consistent data, the measure of openness
used in this research is total trade over GDP, which is total imports plus total exports as a
percent of per capita GDP. This measure was obtained from the Penn World Tables. As
the openness variables are different, their correlation values are not that high: 0.1905 with
the BD openness variable and 0.2517 with the ELR openness variable.
Another measure of openness considered and rejected due to its low correlation with
growth was the total domestic trade as a percent of world trade. The data for world trade
27
for this variable was obtained from the WTO Total Merchandise Trade data set, while
world GDP data came from the World Bank World Development Indicator Database,
2013. Like the trade as a percent of GDP openness variable, this openness variable is not
all that similar to the Sachs-Wagner openness variable used by BD and ELR. Thus the
correlation values are relatively low, at 0.0754 and 0.1319, respectively.
The average country in the average year had trade equal to 57.66% of their GDP. The
median trade value was 52.30%, suggesting a more even distribution of this variable. The
standard deviation of 32.37 means that 95% of the sample observations should fall
between ±64.74% of the mean, or between 0% and 122.40%. The minimum sample
observation was 5.18%, while the maximum was 205.54%.
Policy Variable
As mentioned previously, the policy variable is an index of three variables,
constructed for each time period or sample set. The variables that are included in this
index are the balanced budget or budget deficit variable, the inflation variable, and the
openness variable. To create this variable, the original BD “Policy Index Forming”
regression was estimated:
growtht,i = β0 + β1lgdpt,i + β2ethnolinguistic_fractionalizationt,i + β3SSAt,i +
β4EAsiat,i + β5 institutional_qualityt,i + β6(M2/GDP)t-1,i +
β7budget_surplust,i + β8inflationt,i + β9opennesst,i + ε
(2)
Individual sample values for the balanced budget, inflation and openness variables are
then used to determine the policy value for that sample observations, such that:
28
Policyt,i= β0 + β7budget_surplust,i + β8inflationt,i + β9opennesst,i
(3)
In this research, there are three different time period and two different samples that
require the construction of separate policy variables. The separate time periods are the
original BD time frame of 1970-1993, the original ELR timeframe of 1970-1997, and the
full timeframe of all available data. For the first sample, these three policy variables
contain data for all ELR countries for which data is available. The second sample is for
the only the original BD countries, in the BD analysis years of 1970-1993. Additionally,
the interest in two different inflation variables necessitates the construction of a second
set of policy variables for each time period and sample, one using the CPI inflation
variable as the inflation variable in the policy index-forming regression, and the other
using the GDP deflator.
For the differences with the openness variable and the inclusion of a different
inflation variable, the correlation values between the original BD and ELR policy
variables and the policy variables generated through this research were surprisingly high.
The BD policy variable correlation value with the CPI policy variable was 0.6519, while
that with the GDP deflator variable was 0.6787. The ELR policy variable correlation
value with the CPI policy variable was 0.7743, while its correlation with the GDP
deflator was 0.8028. Further, the CPI policy variable correlation value to growth was
0.3417, while the GDP deflator value was slightly weaker at 0.3369.
The mean of the CPI policy variable is 3.93, while the median is 4.07. The standard
deviation is 0.77, indicating that 95% of all sample observations should fall within ±1.54
29
of the mean, or between 2.39 and 5.47. However, the sample minimum is -3.58 and the
maximum is 6.15.
The mean of the GDP deflator policy variable is 6.17, with a median of 6.31. The
standard deviation for this variable is 0.79, indicating that 95% of all sample observations
should fall between ±1.58 of the mean, or between 3.01 and 9.33. The sample minimum,
however, is -0.52, while the maximum is 8.90.
Initial GDP
The GDP variable used in the original works is the natural logarithm of per capita
GDP for the first year in each time period in constant 1985 US$. This data was retrieved
from the Penn World Table, 1991. For this work, two values of initial GDP were used.
The first was the natural log of per capita GDP for the first year in each time period in
constant 2000 US$. This data was retrieved from the World Bank World Development
Indicators Database, 2013. Missing data was also located at the United Nations Statistics
Division, National Accounts Division of Main Aggregates10, or calculated from per
capita GDP growth values11. As compared to the data in the original studies, the
economic growth data in this study has a 0.8976 correlation value with the economic data
in both the BD and the ELR studies.
The mean value for this measure of initial GDP is $1623.56 per capita. Thus, the
average citizen in the average sample country could expect an income of $1623.56 in a
given year between 1962 and 2009. With a standard deviation of $1806.14, this indicates
10
11
Myanmar 1970-2010: values were extracted in current US$ and converted into 2000US$ using the CPI
Myanmar 1960-1969
30
that 95% of all sample countries should fall between ±3612.08 of the mean, or within the
range of $0 and $5235.84 per capita. The median per capita GDP value for the sample
was $953.37 and no country actually had negative GDP, suggesting that there were many
countries with years of lower levels of per capita GDP.
The second value of initial GDP used in this study was the natural logarithm of per
capital GDP for 1970. The data for this variable was also retrieved from the World Bank
World Development Indicators Database, 2013. Missing data was retrieved from the
Penn World Tables12.
The average for this value is $1317.37, while the median is $776.23. As with the
original variable for GDP, the data still suggests that most countries have a smaller per
capita GDP, while a few have a much larger per capita GDP. With a standard deviation of
$1393.17, the range within which 95% of all sample data should fall is between ±
$2786.34, or within the range of $0 and $4103.71. The minimum 1970 GDP was $121.24
for Malawi, while the maximum was $6611.23 for Argentina.
Institutional Quality
The institutional quality variable is a measure of political risk at a particular point in
time, 1980 for BD and 1982 for ELR. The basis for this measure comes from Knack and
Keefer (1995), who argue that an economy needs a stable, cooperative political
environment to grow. ELR use an average of the PRS Group’s IRIS III data set scores
for: the expropriation risk, the rule of law, the government credibility rating, corruption in
12
Ethiopia, Haiti, Jordan, Tanzania, Uganda
31
government, and the quality of bureaucracy. The higher the value, the better is the
institutional quality. This research uses the 1982 average that ELR use in their study,
which has a 0.0975 correlation with growth.
The mean institutional quality score was 4.36 out of 10, while the median score was
4.48, suggesting that more countries in the sample scored high, but those that scored
poorly, did much worse than those that scored well. The standard deviation was 1.56,
suggesting that 95% of all sample should have ICRGE scores that fall between ± 3.12 of
the mean or between 1.24 and 7.48.
Lagged money supply
The money supply variable is the M2 definition of money, which includes all physical
cash and coin, as well as most easily accessible deposits and funds. The variable is
expressed as a percent of GDP and lagged one time period. The same definition of money
supply is used for all three studies. Both ELR and this research obtain this data from the
World Bank World Development Indicator database, 2002 and 2013, respectively.
Missing data for this research was filled using data from ELR13.The variable in this
research has a correlation value of 0.7839 with the BD money supply variable, and
0.8047 with the ELR money supply variable.
The mean value for the money supply was 43.21% of GDP, thought the median value
was 27.90% of GDP. This suggests that most countries in most years had smaller money
supplies than the average, but those with a larger money supply had a substantially larger
13
Observation (country and time period) from this source were: Democratic Republic of the Congo 8; Haiti 1-6
32
one. Supporting this is the very large standard deviation value of 255.86. Money supply
in the sample does not become negative. So 95% of all money supply values in the
samples should fall between ±511.72 of the mean, or 0 and 554.93% of GDP. The
minimum is 4.16, while the maximum is 6939.22.
Regional Dummies
Regional Dummy variables are used to account for similarities between countries in a
given region that affect the dependent variable. There are five regional dummy variables
in this research. Two, Sub-Saharan Africa (SSA) and East Asia (EAsia), are used in the
primary regressions. SSA are all the countries in the sample that are located in SubSaharan Africa. EAsia are the countries in the sample BD and then ELR considered to be
fast growing East Asian economies: Indonesia, South Korea, Malaysia, the Philippines,
and Thailand. The other three, Egypt, Central America (CentAm), and French Zone (Frz)
are instrumental variables in the two-stage least squares (2SLS) regressions to instrument
for aid and aid*policy. As these are dummy variables and not open to much ambiguity or
interpretation, they are the same for all three studies. In this research, the SSA variable
has a -0.1980 correlation value with growth. EAsia has a 0.1841 correlation value with
growth. Egypt, CentAm and Frz have growth correlation values of -0.1244, -0.0356, and
-0.1160, respectively.
33
Time Period Dummies
Like regional dummy variables, time period dummy variables are used to account for
similarities across the sample, at a given point in time, that affect the dependent variable.
As there are three different data sets for this research, each has a different number of time
period dummy variables. However, like the regional dummy variables, since the time
periods in this research are the same as the time periods used in the BD and ELR but for
the addition of new time periods in the final data set, they are unambiguous and the same
as the BD and ELR sample variables. In the full dataset, 1966-2009, the 5th time period
(1982-85) has the highest absolute growth correlation at -0.2157, followed by the 2nd
(1970-73) and 11th (2006-09) time periods with growth correlation values of 0.1218 and
0.1122, respectively. The 8th (1994-97) and 1st (1966-69) time periods have the lowest
growth correlations values, at 0.0304 and 0.0478, respectively.
Assassinations and Ethnolinguistic Fractionalization
The assassination and ethnolinguistic fractionalization (Ethnf) variables were both
used in the original BD and ELR research, as well as an interaction term between the two.
The ethnolinguistic fractionalization variable is a measure of how fractured the language
structure of the country is by estimating the likelihood that two individuals will belong to
different ethnic groups. If they do, the implication is that they are likely to speak different
languages and have more trouble conducting economic transactions, thus slowing the
mobility and growth potential of the economy. Likewise, assassinations measure the
number of political assassinations per 100,000 of the population and are a proxy for the
34
political stability of the government. Neither of these variables, nor their interaction term,
was significant in any of the original analyses, and current assassination data was not
available. Thus, the Assassination variable and the interaction term were not used in this
research. The ethnolinguistic fractionalization term is a country specific constant that
ELR borrowed from Easterly and Levine (1997), and that is the Ethnf variable selected
for use in this research as well. It has a -0.0810 correlation value with growth.
Population
Although not used directly in any regression, the population variable is used as an
instrumental variable for aid and aid*policy in the 2SLS regressions conducted in this
and the ELD and BD research. The population variable used in all three datasets is the
natural logarithm of total population. For both the ELR data and the data used in this
research, the information came from the World Bank World Development Indicator
database, in 2002 and 2013, respectively. The BD population data and the data for this
research have a correlation value of 0.9997, while the ELR population data and the data
for this research have a correlation value of 1.000.
The mean natural logarithm of population is 16.24, with a median of 16.17. The
standard deviation for this variable is 1.42, indicating a normal range of ±2.84 from the
mean, or between 13.40 and 19.08. The minimum value for this study is 12.90, while the
maximum is 20.89.
35
Lagged Arms imports as percent of total imports
Like the population variable, the arms import variable is not used directly in any
regression, but as an instrument for aid and aid*policy in this and both previous works.
This variable measures total arms imports as a percent of total imports, lagged one time
period. The original BD data came from the World Bank. ELR retrieved their data from
the US Department of State and using the World Bank as a backup resource. The data for
this research primarily came from the World Bank World Development Indicator
database, 2013. However, missing arms imports data was retrieved from the UNCTAD
merchandise trade matrix, and missing total import data was retrieved from the
UNCTAD value of merchandise imports matrix14. Other missing data was included from
the original ELR data15. This variable has a 0.6142 correlation value with the BD arms
import variable, and 0.6009 with the ELR arms imports variable. It has a -0.0567
correlation value with growth in this research, but 0.3439 with aid, 0.3279 with
aid*policy (CPI) and 0.3262 with aid*policy (GDP deflator).
Arms samples from this source were: Bolivia 1996-98, 2001-02; Botswana 1999, 2005-08; Burkina Faso 1995-2004, 2008; Cameroon
1995, 2000, 2003-04, 2007, 2009; Democratic Republic of the Congo 1995, 2003-05, 2007; Republic of the Congo 1995-96, 1998, 2000-04,
2007-08; Costa Rica 1995-96, 1998, 2000-2004, 2007-08; Côte d’Ivore 1996-98, 2000-01, 2005-10; Dominican Republic 1995, 1997, 2002,
2006-08; Ecuador 2000, 2003; El Salvador 1997, 1999-2001, 2004-2007, 2009-10; Ethiopia 1995-96, 2001, 2005-09; Gabon 1995-2003,
2005, 2008-09; Ghana 1995, 1998, 2005-06, 2008, 2010; Guatamala 1996-97, 1999-2010; Guyana 1995-2000, 2002-2010; Haiti 1995-2010;
Honduras 1995-2000; Jamaica 1995-2010; Kenya 1995-96, 1998-99, 2001-06, 2008-09; Madagascar 1995, 1997-2010; Mali 1995-96, 1999,
2001, 2003-04, 2006, 2010; Morocco 1998, 2003-04; Nicaragua 1995-2010; Niger 1995-96, 1998-2002, 2004-07, 2010; Nigeria 1998-99,
2005; Papua New Guinea 1995, 1998-2010; Paraguay 1995, 1997-2000, 2002-03, 2006-09; Philippines 1999-2000; Senegal 1997-2004, 2010;
Sierra Leone 1996, 2000-05, 2007-10; Sri Lanka 2009; Syria 2007; Tanzania 1995, 1997-2002, 2004, 2007-08, 2010; Togo 1996, 1998-2003,
2005-09; Trinidad & Tobago 1995-99, 2003-06, 2008-09; Tunisia 2000-01, 2004, 2007, 2009; 1996-97, 2001, 2003, 2007; Uruguay 2001,
2003-04; Zimbabwe 1995-97, 2002, 2004, 2007-10.
Total imports samples from this source were: Bolivia 1962-69; Ethiopia 1960-80; Iran 1960-64, 2008-10; Jordan 1960-75; Mali 1960-66,
2008-10; Papua New Guinea 2005-10; Sri Lanka 1960-83; Syria 1960-74; Tanzania 1960-89; Trinidad & Tobago 2009-10; Turkey 1960-86;
Uganda 1960-81; Venezuela 1960-73; Zimbabwe 1960-75.
This data was available in current dollars. The CPI was used to convert it to constant 2000$.
15 Observation (country and time period) from this source were: Botswana 1, 2, 5; Burkina Faso 6; Cameroon 7; Democratic Republic of
the Congo 7; Republic of the Congo 7; Costa Rica 7; Côte d’Ivore 5-7; Dominican Republic 6, 7; El Salvador 6; Ethiopia 6; Gambia 1-7;
Ghana 1, 5; Guyana 6; Haiti 1, 6, 7; Honduras 1, 2, 7; Jamaica 1, 4; Madagascar 3, 6, 7; Mali 2, 6; Nicaragua 1, 7; Niger 2, 3, 6, 7; Papua
New Guinea 2; Senegal 6, 7; Sierra Leone 1, 2, 4, 6; Tanzania 5, 6; Togo 6; Trinidad & Tobago 1, 3, 5-7; Uganda 4, 5, 7; Zambia 6, 7;
Zimbabwe 1.
14
36
The mean arms1 variable is 3.32% indicating that in a given year, the average country
imported 3.32% of GDP worth of arms. However, the median of 0.98% suggests that
most countries import smaller portions of their GDP, while a few import greater values of
arms. The standard deviation is 7.61%, suggesting the range in which 95% of all sample
observations should fall is between -11.9% and 18.54% of GDP. The dataset minimum is
-8.54%, while its maximum is 85.01%.
Interaction Variables
An interaction variable is one that combines two or more variables to make a new
variable, allowing researchers to test variable relationships to see if certain variables have
sum effects greater than the sum of their individual parts. In this research, the interaction
variable contains the policy variable with some other variable. With the exception of the
aid*policy terms, all of the interactive variables are used exclusively as instrumental
variables by BD and ELR to estimate aid and aid*policy in their 2SLS regressions. In all,
there are four sets of interaction variables. They are: aid*policy and aid2*policy,
arms1*policy, lpop*policy and lpop2*policy, and lgdp*policy and lgdp2*policy.
Aid*Policy and Aid2*Policy
Both policy variables have relatively low correlations with growth for this interaction
relationship. For the CPI policy variable, the aid*policy variable has a 0.0000 correlation
with growth, and the aid2*policy variable has a 0.1020 correlation with growth. For the
37
GDP deflator policy variable, the aid*policy variable has a -0.0056 correlation with
growth, and the aid2*policy variable has a 0.1068 correlation with growth.
Instrumental variables
None of the instrumental variables have an exceptionally strong correlation with
growth, which is ideal. However, some are much higher than others. They are also similar
between the two policy variables.
The arms1*policy interaction term has -0.0423 (CPI) and -0.0734 (GDP deflator)
correlation with growth. The correlation with aid is -0.1929 (CPI) and -0.2004 (GDP
deflator). The corralation with aid*policy is -0.1193 (CPI) and -0.1276 (GDP deflator).
The LGDP2*policy interaction term, which is the natural logarithm of (1+ the square
of the 1970 GDP), has 0.1408 (CPI) and 0.1194 (GDP deflator) correlation with growth.
The correlation with aid is -0.3211 (CPI) and -0.3688 (GDP deflator). The correlation
with aid*policy is -0.2726 (CPI) and -0.3341 (GDP deflator).
The population*policy interaction term has 0.3399 (CPI) and 0.3410 (GDP deflator)
correlation with growth. The correlation with aid is -0.1829 (CPI) and -0.2438 (GDP
deflator). The correlation with aid*policy is -0.3093 (CPI) and -0.3049 (GDP deflator).
The population2*policy interaction term has 0.3238 (CPI) and 0.2898 (GDP deflator)
correlation with growth. The correlation with aid is -0.2050 (CPI) and -0.2743 (GDP
deflator). The correlation with aid*policy is -0.1154 (CPI) and -0.2203 (GDP deflator).
38
The LGDP*policy interaction term has 0.2600 (CPI) and 0.2377 (GDP deflator)
correlation with growth. The correlation with aid is -0.2504 (CPI) and -0.3180 (GDP
deflator). The correlation with aid*policy is -0.1346 (CPI) and -0.2641 (GDP deflator).
These correlation values are important as the rule of thumb instrument selection
suggests that good instruments have a weak relationship to the dependent variable, but a
strong relationship with the variable they are to instrument.
Low-Income Variable
The low income variable was developed in the BD study to denote the low-, as
opposed to medium-income, countries so that a set of regressions could be conducted
using a subset of the sample. They assign low-income status to those countries with real
per capita GDP below $1900 in 1970. Their one exception to this was Nicargua, which
had started out at a higher income level, but for whom real per capita GDP had fallen
below $1900 by 1982. This variable was not, however, ever used as a control or variable
of interest. ELR changed this variable from a positive to a negative, but otherwise left it
the same and used it in the same fashion. As a quasi-dummy variable, as the values are -1
or 0, the variable is not particularly ambiguous. One could dispute the definition of lowversus middle-income. However, all of the low-income countries were classified as at
least lower-middle income countries by the OECD at the time of the BD paper16, and
only Botswana had risen to an upper-middle income country by the time of the ELR
paper’s publication17. Before any countries are dropped due to insufficient data in
16
17
OECD DAC List of Aid Recipients Used for 1997, 1998 and 1999 Flows
OECD DAC List of Aid Recipients- As of 1 January 2003
39
particular regressions, there are 18 middle-income and 46 low-income countries, and the
variable has a -0.0376 correlation with growth. When the variable is switched so that the
low-income countries are represented by 1 and the middle income countries by 0, the
variable growth correlation becomes -0.0443.
Outlier Variable
The outlier variable is assessed from the aid*policy interaction variable, as this is the
variable of greatest interest in the original BD research. To determine the outliers, the
mean and standard deviation of the aid*policy variable must be determined. Next, the
parameters for two standard deviations above and below the mean need to be established.
Once that has been accomplished, any aid*policy sample that is outside of those ±two
standard deviation range is considered an outlier. As the openness variable of the policy
index was different from the original BD and ELR data, the outlier variable has similar
differences, skewing to higher more positive integers, where the BD and ELR variables
were more evenly distributed, positive and negative. As a result, the BD outlier has a 0.0172 correlation value with the outlier value for this research, while the ELR outlier has
a -0.0245 correlation value with the outlier value for this research. In this research, the
outlier has a -0.1917 correlation with growth. However, like the low-income variable, the
outlier variable is not used in the BD or ELR research as a variable of interest or a control
variable. Rather it is used as a sample control variable, to exclude outliers from particular
regressions.
40
Chapter 4
EMPIRICAL METHODOLOGY
The literature explaining how foreign aid fits into the economic growth and
development models remains inconclusive at best, while seeming confusing and
inconsistent at other points. The models draw from a wide swath of theoretical
applications. Aside from economic models of growth, these models also draw on sociopolitical, historical and geo-political elements including political structure and
government type, region and physical attributes of the country, religion and language,
former colonial power and length of independence. This has contributed to three basic
groups of theory. The first is based primarily on the economic growth theory which
derives growth from economic inputs of labor and capital. The next follows the
macroeconomic theory that a nation’s income is equal to private consumption,
government spending, investment and net exports; thus, economic growth is derived from
changes in these variables. The third group primarily bases economic growth on the
previously mentioned geopolitical, social and historical variables. Many theories,
however, borrow from each of these groups, so there are no clearly defined schools of
thought as to how aid influences economic growth.
Further, there are several different theories on how aid goes about impacting
economic growth. The basic idea is that if you add resources to an economy, it will have
higher levels of output and that higher level of output will constitute a higher level of
growth. However, our basic economic growth theory would tell us this growth is just a
one-time infusion that that would not be carried over into proceeding time periods. Since
41
the role of most aid is to improve living standards and alleviate poverty, it is consumed
and thus only contributes to single time period growth. However, a portion of aid is
intended for investment, improved knowledge, infrastructure and technology. This type
of aid will allow long-run growth without any additional allocation.
Fortunately for most poor countries, they do receive aid allocations annual, even
though the amount of those allocations may not be reliable. This research asks the
question of how developmental aid impacts annual economic growth. To do that, the
model and tests first used by BD and then ELR will be used. In order to test the
relationship of aid to growth, and the role that policy might play in the relationship, the
following equation is estimated using a panel dataset:
growth = ƒ[initial pcGDP, ethnic fract, inst quality, m2/GDP (lagged), SSA,
EAsia, policy, aid/gdp, aid*policy, aid2*policy]
(1)
Before this can occur, the estimation of (1) requires the construction of a variable that
measures the quality of policy for each sample observation, or each country in each time
period. This policy construction follows the procedure described in BD:
The policy variable consists of direct influences from the budget surplus, inflation,
and openness variables. In order to estimate this, these variables are estimated in the
growth equation without the aid and policy variables so that:
growtht,i = β0 + β1lgdpt,i + β2ethnolinguistic_fractionalizationt,i + β3SSAt,i+
β4EAsiat,i + β5institutional_qualityt,i + β6(M2/GDP)t-1,i +
β7budget_surplust,i + β8inflationt,i + β9opennesst,i + ε
(2)
42
Construction of the policy variable is then done using the estimated coefficients:
Policyt,i = Constant + β7budget_surplust,i + β8inflationt,i + β9opennesst,i
(3)
As there are two different samples and three time periods, a separate policy variable
needs to be constructed for each. Further, in this research, there are two different inflation
variables, CPI and GDP deflator, to be tested. So two sets of policy variables need to be
constructed for each sample and time period within the scope of this research.
Next, once the policy variables have been estimated, the aid-policy interactive terms
can be constructed. From the aid*policy variable, the outlier variable is determined using
the Hadi (1992) method of classifying observations ±2 standard deviations away from the
mean as outliers. Now each set of tests conducted has unique policy, aid-policy
interactive and outlier variables.
Finally, it is possible that there may be endogeneity between growth and aid. If there
is, aid and any interactive variables need to be instrumented. The instruments that BD
determined to be the best instruments have the following relationship with aid:
(aid, aid*policy) = ƒ[central america, egypt, french zone, arms imports/total
imports (lagged), log population, arms1*policy, lpop*policy,
lpop2*policy, log initial gdp*policy, lgdp2*policy]
(4)
Within each data set that is tested, ELR follow a specific path in their quest to test the
robustness of the BD findings that good policy is necessary for aid to contribute
positively to growth. First they estimate (1) with outliers, calling this 4/OLS.
43
growtht,i = β10 + β11lgdpt,i + β12ethnolinguistic_fractionalizationt,i + β13SSAt,i
+ β14EAsiat,i + β15institutional_qualityt,i + β16(M2/gdp)t-1,i + β17aidt,i +
β18aid*policyt,i + β19aid2*policyt,i
(4/OLS)
Next they estimate (1) without the aid2*policy variable, without and then with
outliers, calling these 5/OLS and 5+/OLS, respectively.
growtht,i = β20 + β21lgdpt,i + β22ethnolinguistic_fractionalizationt,i + β23SSAt,i +
β24EAsiat,i + β25institutional_qualityt,i + β26(M2/gdp)t-1,i + β27aidt,i +
β28aid*policyt,i
(5/OLS)
growtht,i = β29 + β30lgdpt,i + β31ethnolinguistic_fractionalizationt,i + β32SSAt,i +
β33EAsiat,i + β34institutional_qualityt,i + β35(M2/gdp)t-1,i + β36aidt,i +
β37aid*policyt,i
(5+/OLS)
The following steps re-test 5 and 5+ using the 2SLS instrumental variable method,
without and then with outliers, calling these 5/2SLS and 5+/2SLS, assuming there is
endogeneity between the aid variables and growth, but without testing to verify the
appropriateness of this estimation or then the strength of the instruments.
growtht,i = β38 + β39lgdpt,i + β40ethnolinguistic_fractionalizationt,i + β41SSAt,i +
β42EAsiat,i + β43institutional_qualityt,i + β44(M2/gdp)t-1,i + β45aidt,i +
β46aid*policyt,i
(5/2SLS)
growtht,i = β47 + β48lgdpt,i + β49ethnolinguistic_fractionalizationt,i + β50SSAt,i +
β51EAsiat,i + β52institutional_qualityt,i + β53(M2/gdp)t-1,i + β54aidt,i +
β55aid*policy
(5+/2SLS)
44
This sequence of regressions is repeated for a restricted low-income country subset of
the sample for estimations 7/OLS, 8/OLS, 8+/OLS, 8/2SLS, and 8+/2SLS.
Focused on the issues of aid and aid*policy robustness, the ELR methodology does
not test for endoegeneity, the significance of outliers or low-income countries,
multicollinearity, or any other potential problem. BD did test for endogeneity and
determined that it was not a problem within their result. They also compare coefficient
estimations between the full sample results and those that were generated for the lowincome countries, concluding that there is no difference between how aid and policy
impact growth for the low-income versus full sample of countries. They do not, however,
test this conclusion. Further, both studies conclude that there is not a significant
difference between results estimated with and without outliers through observation, but
do not actually test this conclusion.
The estimation of these equations in this research is performed using a panel
dataset with 63 countries and three (3) time samples covering the time period 1966-2009,
including time dummy variables.
45
Chapter 5
ANALYSIS
Estimations: Burnside-Dollar Countries and Years (1970-1993)
See the Results Tables 1 and 2 in Appendix B for CPI and GDP Deflator Policy for
the BD Countries and Years to see a complete accounting of estimation results. In the
first estimation in this set of regressions, the policy index-forming regression, the budget
surplus coefficient was positive and significant to the 99% confidence level for both the
CPI policy index and the GDP deflator policy index. However, neither had significant
inflation (-) or openness (+) coefficients, though they were of the expected sign. In the
estimations that follow, the aid, aid*policy, and aid2*policy coefficients are never
statistically significant for either policy index, and switch sign. However, aid most often
has a positive coefficient and aid*policy most often has a negative coefficient. The
aid2*policy coefficients are always positive.
Of the variables of interest, policy is the only variable that is significant and has a
consistent sign. For the GDP deflator policy index, policy is positive and significant for
every estimation except 8/2SLS, when the low-income country sub-sample is estimated
excluding outliers using instrumental variables for aid and aid*policy. Since policy is
significant in the full sample, this suggest that there is something different about this subsample that might make policy less important to growth than it is to the full sample.
For the CPI policy index, policy is positive and statistically significant except for the
regression mentioned above, 8/2SLS, as well as 5/OLS (full sample, no outliers), 5/2SLS
(full sample, instrumented aid variables, no outliers), 5+/2SLS (full sample, instrumented
46
aid variables, outliers included), and 8/OLS (low-income sample, no outliers). In all other
cases, policy is significant to at least the 95% level of confidence.
When tested, both sets of estimations did not have problems with endogeneity or
omitted variables. The White test confirmed heteroskedasticity in both, and robust
standard errors were used. High levels of multicollinearity were detected only between
the aid and aid*policy variables. Since one is an interactive term including the other, it is
not surprising or particularly concerning.
Estimations: Easterly-Levine-Roodman Countries and Burnside-Dollar Countries Years
(1970-1993)
See the Results Tables 3 and 4 in Appendix B for CPI and GDP Deflator Policy for
the ELR Countries and BD Years to see a complete accounting of estimation results. In
the first estimation in this set of regressions, the policy index-forming regression, the
budget surplus coefficient was positive and statistically significant for both the CPI
policy index and the GDP deflator policy index. Further, the inflation variable for each
was negative and statistically significant. However, neither had a significant openness
coefficient, though it was of the expected sign.
In the estimations that follow, for both policy indices aid2*policy coefficients are
always positive and statistically significant to the 99% confidence level. However, none
of the other aid or aid*policy variables were significant and sometimes they switch signs
in the GDP deflator policy index estimations, although they are mostly positive. In the
CPI policy index estimations, aid only has significance in the 5/2SLS and 5+/2SLS
47
estimations. Aid*policy only has significance in the 5+/2SLS and 7/OLS estimations,
although the coefficient switches signs between these results.
Again, for both sets of estimations, the policy index is positive and statistically
significant for the majority of the estimations. For the GDP deflator, the only estimations
where policy is not significant to at least the 95% confidence level are 8/2SLS and
8+/2SLS. For the CPI policy index, the only estimations where policy is not singificant
are 8+/OLS and 8+/2SLS.
When tested, both sets of estimations did not have problems with endogeneity or
omitted variables. The White test confirmed heteroskedasticity in both, and robust
standard errors were used. High levels of multicollinearity were detected only between
the aid and aid*policy variables. Since one is an interactive term including the other, it is
not surprising or particularly concerning.
Estimations: Easterly-Levine-Roodman Countries and Years (1970-1997)
See the Results Tables 5 and 6 in Appendix B for CPI and GDP Deflator Policy for
the ELR Countries and Years to see a complete accounting of estimation results. In the
first estimation in this set of regressions, the policy index-forming regression, the budget
surplus coefficient was positive and statistically significant to the 95% level of
confidence for both the CPI policy index and the GDP deflator policy index. Further, the
inflation variable for each was negative and statistically significant to 95%. However,
neither had a significant openness coefficients, though it was of the expected sign.
48
In the estimations that follow, for both policy indices aid2*policy coefficients are
always positive and statistically significant to the 99% confidence level. The only other
aid variable that has any significance for the GDP deflator policy index estimations are
the aid*policy coefficients for the 5/2SLS, 5+/2SLS, and 8/2SLS, where they are positive
and statistically significant to the 90% level of confidence. In the CPI policy index
estimations, aid is positive and significant to the 95% confidence level for the 5+/OLS
and 8+/OLS coefficients. For aid*policy, the coefficients for 4/OLS, 5+/OLS, 7/OLS,
and 8+/OLS are all negative and statistically significant to at least the 90% confidence
level.
Policy is again positive for all of the estimated coefficients in both indexes, and
statistically significant for most. For the GDP deflator policy index estimations, the
coefficients are not significant for any of the 2SLS estimations. For the CPI policy index
estimations, the coefficients are not significant for the low-income 2SLS regressions,
8/2SLS and 8+/2SLS.
When tested, both sets of estimations did not have problems with endogeneity or
omitted variables. The White test confirmed heteroskedasticity in both, and robust
standard errors were used. High levels of multicollinearity were detected only between
the aid and aid*policy variables. Since one is an interactive term including the other, it is
not surprising or particularly concerning.
49
Estimations: Easterly-Levine-Roodman Countries and All Years (1966-2009)
See the Results Tables 7 and 8 in Appendix B for CPI and GDP Deflator Policy for
the ELR Countries and All Years to see a complete accounting of estimation results. In
the first estimation in this set of regressions, the policy index-forming regression, the
inflation variable was negative and statistically significant to 99% level of confidence for
both the CPI policy index and the GDP deflator policy index. For the CPI policy index,
the budget surplus coefficient was positive and significant to the 90% confidence level,
while for the GDP deflator policy index it was negative and significant to the 99%
confidence level. However, neither had a significant openness coefficients, though it was
of the expected sign.
In the estimations that follow, for both policy indices aid2*policy coefficients are
always positive and statistically significant to the 99% confidence level. However, none
of the other aid or aid*policy variables as significant and sometimes switch signs in the
GDP deflator policy index estimations. In the CPI policy index estimations, aid only has
significance in the 5/2SLS and 8+/OLS estimations, where they are positive and
significant to the 95% confidence level. Aid*policy only has significance in the 8+/OLS
estimation, where it is negative and significant at the 95% confidence level.
Policy is positive and statistically significant for all estimations in both the GDP
deflator and CPI policy indexes.
When tested, both sets of estimations did have a problem with endogeneity in both
the aid and aid*policy variables, but there was no omitted variable bias. The White test
confirmed heteroskedasticity in both, and robust standard errors were used. High levels of
50
multicollinearity were detected only between the aid and aid*policy variables. Since one
is an interactive term including the other, it is not surprising or particularly concerning.
Conclusions from these Estimations
First, policy, either determined by CPI or GDP deflator as the inflation variable in the
policy index, has a strong positive relationship with growth.
Second, aid and aid*policy do not seem to have a significant relationship with
growth. However, since the aid2*policy variable is almost always positive and
statistically significant (in every set of estimates except the BD countries and years), this
suggest that there is some kind of relationship. This may make it difficult to estimate a
relationship between growth and aid when aid and aid* policy are included due to
multicollinearity. Since, aid*policy is an interactive term of aid, it is expected that they
would be collinear. When the test was conducted for each set of estimations,
multicollinearity between aid and aid*policy was confirmed. Thus it is possible that they
are significant to the growth estimation. However, determining the level and direction of
impact for each with a high level of certainty is unlikely.
Third, while these results show a similar lack of robustness to the ELR results, ELR
did not test the data or models to see if there are any problems that might influence
robustness. They did not test for endogeneity in their estimations. Had they done so, they
could have said whether the OLS or 2SLS were the appropriate estimation method and
tested for any other estimation-specific problems, like weak instruments in 2SLS
estimations. Had they tested the outliers or low-income countries to see if they were
51
statistically different from the rest of the sample, they could have said if the estimations
that excluded outliers or middle-income countries were statistically different. Yet, they
did none of these things. Rather, they estimated several sets of aid and aid*policy
coefficients, saw that they vary in sign and significance, and concluded that the aidpolicy relationship to growth must not be robust.
Fourth, these results show that the East Asia regional dummy and institutional quality
variables are positive and almost always statistically significant in detremining growth.
Likewise, the Sub-Saharan Africa and initial GDP variables are negative and most often
statistically significant in determining growth. The money supply is more inconsistent
across sets of estimations, but within estimations groups, if it is significant, it is negative,
with a coefficient of 0.001, and significant to the 99% confidence level.
Finally, it seems like the best results from these results are the full time period
estimation of 5/2SLS in both the GDP deflator and CPI policy index estimation sets.
Tests, explained in the next section, conclude that there is endogeneity (2SLS is relevant
over OLS estimation), find that there is a significant difference between the outlier and
non-outlier sample observations (outliers are excluded), and do not find a significant
difference between low- and middle-income countries (full sample for larger sample size
and broader application of results). Between the two policy indexes, the estimated
coefficients are somewhat different in scale throughout the whole set, but tend to move in
the same direction and to the same degree. For this particular estimation, the CPI policy
index aid variable is negative and statistically significant to the 95% level of confidence.
52
For the GDP deflator index it is also negative but not significant. This could, however, be
explained by multicollinearity.
Further tests are then performed on these two estimations to find the estimation model
that best represents the aid-policy-growth model and answers the question of how
developmental aid, and policy through aid, impact economic growth.
53
Chapter 6
RESULTS, ROBUSTNESS TESTS, AND INTERPRETATION
Final Estimation Details
To determine a final specification for the growth regression, an extension is made
from the original Easterly, Levine, and Roodman and Burnside-Dollar regressions, by
extending the dataset to include all. The final specification includes a longer timeframe
and the full sample of developing countries, not just those with low levels of per capita
GDP or growth. Variables that fail to exhibit significance throughout any of the
specifications are dropped. Finally, tests are conducted and changes will be made to
correct for heteroskedasticity, autocorrelation, omitted variables, sample breaks, and
endogeneity.
The most representative regression for this research question is as follows:
growtht,i = -6.02 + 65.59Aidt,i – 0.25Aid*Policyt,i – 56.28Aid2*Policyt,i
(2.28) **
(-0.06)
(-2.00) **
– 0.001M2t,i + 1.09Policyt,i
(-4.94) ***
(4.45) ***
( Panel OLS)
In total, 532 observations were included in this regression, for 61 countries, across
eleven four-year time periods, starting in 1966, and ending in 2009. This was estimated
as a panel OLS regression with fixed effects, excluding outliers and with time period
dummy variables, on the full sample of medium- and low-income developing countries.
As a result, it dropped the variables that are constant within a country, attributing them to
the fixed effect of that particular country. The variables dropped were initial GDP in
54
1970, the Sub-Saharan Africa and East Asia regional dummy variables, and the
institutional quality variable. The variable for money supply, however, was retained and
the coefficient did not change in magnitude or significance from the previous sets of
estimations. For this estimation, there are a minimum of three and a maximum of eleven
observations per country, with an average of 9.1. Within each country group, the R2 is
0.2442, meaning that 24.42% of the change in GDP growth for that country can be
explained by changes in the variables for that country. The overall R2 is 0.0813, meaning
that for the sample as a whole, changes in the variables account for 8.13% of the
observed economic growth. On the surface, this seems like a very small amount.
However, it does not include the impact that the dropped variables might have had on the
whole. Further, it suggests that about 8% of the change in growth across the sample
might be attributed to aid and policy. The F-statistic is 48.76. So we fail to reject the null
hypothesis that the regression as a whole is invalid. This regression was run without
outliers, and includes both middle- and low-income developing countries. As such, the
coefficients for each variable remain about the same, but the statistical significance of the
aid, aid*policy and GDP variables decline from the same regression that does include
outliers.
Each variable impacts growth differently when it changes. Interactive variables, like
aid*policy, influence growth multiple times. When aid increases by 1% of GDP, growth
will increase by 0.66%. This coefficient is positive and statistically significant to the 95%
level of confidence. If the aid and policy variables increase by 1% and one, respectively,
the aid*policy variable increases by 0.01 and growth will decrease by 0.0025%.
55
However, aid*policy is not statistically significant. The same change in the aid and policy
variables will increase the aid2*policy by 0.0001, causing growth to decrease by 0.00563.
This coefficient is statistically significant to the 95% confidence level. When the money
supply as a percent of GDP, lagged one time period, increases 1%, growth will decline by
0.001%. When the policy index variable increases by one, growth increases by 1.09%.
Both the money supply variable and policy are statistically significant to the 99%
confidence level.
As mentioned previously in the initial results section, the aid and aid-policy
interactive terms may be affected by multicollinearity. This is not uncommon for
interactive terms and the original variables. However, it could cause variable to appear
insignificant when in fact they are not, and generally make it harder to determine the
exact relationship between a given independent and dependent variable. When these aid
and aid-policy variables are estimated within the fixed effect panel but independently of
one another, they were all positive. Further, aid and aid*policy were statistically
significant to the 95% level of confidence. Thus it is likely due to multicollinearity that
the aid*policy variable is not significant.
This regression was run with time period dummies. The first time period (1966-69)
was dropped, while the second (1970-73) and third (1974-77) were positive and
statistically significant and the fifth (1982-85), sixth (1986-89), ninth (1998-2001), tenth
(2002-05), and eleventh (2006-09) were negative and significant.
In the original set of estimations, the variables that were dropped by the fixed effects
model had reasonably stable coefficients and significance levels. Since it is the intention
56
of this research to be applicable to the widest range of less developed countries, those
coefficients and their relationship to growth are reported here. If a country’s initial GDP
per capita in constant 2000 US$ in 1970 increased by $1, growth would by about
decreased by 0.535%. If a country is located in Sub-saharan Africa, growth will be
reduced by 1.555%. Conversely, if the country is a fast growing East Asian country,
growth will increase by 1.682%. When the ICRGE country risk index increases by one
point, growth increases by 0.281%.
Evaluations of Robustness
Dropped variables
This regression includes data for all available time periods, to encompass the widest
possible scope of the aid-growth relationship. Similarly, it uses the expanded ELR sample
of countries over the more limited BD sample. From the original regressions,
assassinations, ethnolinguistic fractionalization and assasinations * ethnolinguistic
fractionalization have been dropped (for lack of data on assassinations and) as they are
not significant in the original or any successive regressions.
GDP Deflator
In the initial studies, the policy indices use CPI as the inflation variable, along with
balanced budget and openness variables. In this regression, the GDP deflator was used as
the basis for the inflation variable in the policy index. To test the appropriateness of this
measure in the model, the GDP deflator policy index was tested as the policy variable in
57
the original Easterly, Levine, and Roodman data sets for all countries for 1970-1993 and
1970-1997 with the full set of regressions. In both of the original Easterly, Levine, and
Roodman policy index-forming regressions, the balanced budget variable was not
significant. In the regressions with GDP deflator as the inflation variable, all three policy
index-forming variables were statistically significant to the 99% confidence level.
While in both the original and the alternate (GDP Deflator) regressions, the policy
variable is positive and statistically significant in all regressions, in the original
regressions it is almost always at the 99% confidence level (once at 95%). However, in
the alternate regressions, the confidence level varies from 90-99% confidence. Further
the coefficient for the original policy variable is always higher than the alternate policy
variable, ranging from 0.87- 1.61 for the original policy variable and 0.31-1.10 for the
alternate policy variable.
In every regression in the alternate set of regressions, the aid variable was negative
and the aid*policy variable was positive. They were not always significant, but they were
always consistent in sign. In fact, they are frequently not significant, but they are more
frequently significant than aid or aid*policy in the original regression. In the original
regressions, aid is often positive and aid*policy is often negative, though the signs do
change so that they are the same, positive or negative. For the regressions that also
include aid2*policy, this variable is always positive. We would typically expect it to be
negative, to signify diminishing returns to aid*policy. However, it only has significance
in one of the four regressions in which it is included. Finally, when ELR plot growth
against the aid*policy variable, the result was a negative sloping line. When using the
58
GDP deflator as the inflation variable in the policy variable, this is the result for the same
plot:
Figure 1: Growth-Aid*Policy Plot
Estimated Growth vs Aid*Policy relationship using
the GDP deflator policy index in the original ELR dataset.
Initial GDP
The original initial GDP variable has been changed in this regression from the initial
per capita GDP in 2000 US$ for each time period [this means that, for each 4-year time
period, the GDP was observed at the beginning of the time period and recorded as the
observation for that period] to the initial per capita GDP in 2000 US$ for each sample
country in 1970. Thus, this variable changes from a unstable, potentially cyclical value
that has little to no significance and often exhibiting an unexpected relationship to
growth, to a constant value with a consistently negative relationship to growth, a more
consistent coefficient value and a significance at the 90% level of confidence in many
59
regression specifications. Similarly, when the initial GDP variable was changed, the SSA
variable coefficient became more consistent and significant at the 99% confidence level.
As the initial GDP variable was not statistically significant in all regression
specifications, a test was performed by dropping it from the regressions to see if this
altered the coefficients and their significance; it did. A Ramsey Reset Omitted Variable
test was also conducted before and after dropping the GDP variable, though neither test
result rejected the null hypothesis that the model has omitted variables.
Chow tests for breaks in sample18
It was suspected that there might be a couple of different breaks in the sample. It
seems like there were more inconsistencies with the aid variable in the last group of
regressions, with all time periods, leading to the speculation that there might be a
distinction between aid given during the Cold War and that given after. To test this
hypothesis, a Chow test was performed. For that, two new group variables must be made
within the sample that completely separate, or break apart, the two groups that are
suspected of being different. In the case of the Cold War, Group1 was defined as periods
1-7; so those including 1966-1992. Group2 was defined as periods 8-11; so those
including 1993-2009. Next, each group variable was interacted with all of the
independent variables and growth was regressed against the group1 and group2 variables
as well as all of their interacted variables (aid, aid*policy, ssa, easia, icrge, m21, and
policy). Then, the predicted coefficient for each group1 variable was tested against the
Resource for conducting the Chow test in Stata: http://www.stata.com/support/faqs/statistics/chow-tests/;
updated July 2011; accessed June 2013.
18
60
related group2 predicted coefficient. Individual results showed that none of the
coefficients were statistically significantly different from their related coefficient.
Likewise, when the test was run to accumulate these results, the results were still not
statisticlly significant. Thus we fail to reject the null hypothesis that group1=group2, and
that there is no difference in the effect of aid given during or after the Cold War.
Further, there might be a difference between medium and low income countries, as it
seemed like coefficients and significance for aid, aid* policy, policy, and GDP changed
more for the regressions done with only low income countries than those done with the
whole sample. To test this, the variable that denotes low-income countries was added to
the full-sample regression. However, the coefficient is not statistically significant.
Further, a Chow test was performed in the previously stated method. As before, none of
the individual coefficient tests were statistically different from their related coefficient,
nor are the accumulated results statistically significant. So, again, we fail to reject the null
hypothesis of no difference between the samples, and conclude that there is no
statistically significant difference between middle and low income countries to warrant
distinguishing them in the regression.
Finally, there is a clear distinction in coefficients and significance levels whether
outliers are excluded or included in a regression. As with the test on low and medium
income counties, adding the outlier variable to the full sample regression, where outliers
are not excluded, is performed. The coefficient is not statistically significant. Next, the
appropriate steps are again taken for conducting the Chow test. This time, however, the
results were different. The group2 and EAsia variables are dropped from the estimation,
61
meaning that no coefficient is estimated for these variables. At least for the EAsia
coefficient, this was because there were no East Asia countries in the outlier group. When
the tests of the coefficients are conducted, all of them are insignificant, except the test of
the EAsia coefficient, which is significant to the 99% confidence level. When the
accumulated results were tested, they too were significant to the 99% confidence level.
However, even without the East Asia coefficient, the accumulated difference between the
coefficients result in a p-value that is 0.0510, and we reject the null hypothesis that the
coefficients are the same. If, however, only aid, aid*policy, policy and group are
considered, the p-value is high, we fail to reject the null hypothesis and conclude that the
coefficients of the outliers are not statistically significantly different from the coefficients
of the non-outliers.
Using a predicative tool (dfbeta variable) in Stata, the influence on the coefficient of
excluding the outliers from the sample was estimated. There are twenty-four (24) outliers.
If they are excluded, this tool predicts the coefficient for aid will increase by 0.67, the
coefficient for aid*policy will decrease by 0.58, and the coefficient for policy will
increase by 0.22. If this is accurate, it is suggesting the model is not very robust, if the
exclusion of 24 observations from a total of over 550, can have that large an impact on
the regression results. Thus, the decision was made to exclude outliers.
This test was first performed on the GDP deflator policy index OLS estimations for
all time periods before the determination of endogeneity. Testing of the CPI policy index
OLS estimations for all time periods also yielded similar results. When the final analysis
method of fixed-effect panel OLS estimation was determined, the test was performed
62
again with aid, aid*policy, aid2*policy and money supply as the coefficients of interest.
The test resulted in a p-value of 0.0538, leading to the rejection of the null hypothesis that
the coefficients between the outlier and non-outlier groups are the same and confirmation
that outliers should still be excluded from the estimations.
Ramsey Reset Test
Prior to establishing that outliers were to be excluded, the Ramsey Reset Omitted
Variable test, with an F-stat of 0.97 (p=0.4057) suggests that there is not omitted variable
bias. Additionally, the linktest was conducted to test for other specification errors. No
errors were indicated. When outliers were excluded from the estimations, the results
changed somewhat. The F-stat rose to 1.29 (p=0.2171). However, we still fail to reject
the null hypothesis that the model has no omitted variable bias, and the linktest still
indicates no additional specification error.
With the discovery of endogeneity, the Ramsey/Pesaran-Taylor RESET test was
conducted on the 2SLS estimation results. With a Chi2 of 0.30 and a p-value of 0.5845,
we again fail to reject the null hypothesis and the linktest confirms no specification
errors. However when the decision is ultimately made that the best tool to use with this
data to estimate the relationship between aid-policy-growth is the fixed effects estimation
method, neither the Ramsey RESET nor the linktest for model specification function to
test the model further.
63
Autocorrelation
As this data has been regressed on panel data, with only a short time series element,
autocorrelations is not a strong concern. When estimated with fixed effects in the panel
estimation, the Wooldridge test for autocorrelation in panel data estimated and F-value of
0.620 and a p-value of 0.4341. Thus, we fail to reject the null hypothesis that there is no
first order autocorrelation, and conclude that autocorrelation is indeed not a problem for
this data.
Heteroskedasticity
Heteroskedasticity, however, is a concern. To test for this, the Breusch-Pagan/CookWeisberg is conducted with the original set of Easterly, Levine, and Roodman variables,
minus those previously discussed as having been dropped as insignificant on the original
OLS estimation after outliers were excluded (5/OLS). The initial results for this test was
a Chi2 of 6.21 (p=0.0127), suggesting that heteroskedasticity is present. When the White
test for heteroskedasticity was conducted, the Chi2 of 306.50 and p-vaule of 0.0000
confirm this finding. When tests revealed endogeneity, the 2SLS instrumental variable
estimation results were also tested for heterskedasticity. This test, the Pagan Hall general
test statistic for heteroskedasticity, has a Chi2 of 46.28 and p-value of 0.0085. Thus we
reject the null hypothesis that the disturbance is homoskedastic and again confirming the
presence of heteroskedasticity in the data.
Finally, when the fixed effects panel estimation method is selected over the standard
OLS and GMM estimation methods, the model is again tested for heteroskedasticity. The
64
Modified Wald Test for Groupwise Heteroskedasticity in Fixed Effects estimates a Chi2
of 4245.02 and a p-value of 0.0000. Thus, we again reject the null hypothesis of
homoskedasticity. To correct for this, the regressions were run with robust standard
errors. This does not eliminate the heteroskedasticity, but it limits the extent to which
insignificant variables are perceived as impacting the growth relationship.
Durbin-Wu-Hausman augmented test for endogeneity19
In the original regressions, ELR and BD instrument for aid and aid*policy with a
series of variables that include a dummy variable for Egypt, a dummy variable for
Central American countries, a dummy variable for French Zone countries (former French
colonies or spheres of influence), arms imports as a percent of total imports lagged one
time period (ARMS1), ARMS1* policy, the natural log of population (lpop),
lpop*policy, lpop2*policy, the natural log of GDP(lgdp)*policy, and lgdp2*policy. When
looking at the correlation values for these variables, some interesting relationships
become apparent. The French Zone dummy variable has a fairly low correlation with
growth (-0.12), but is more highly correlated with both aid (0.24) and aid*policy (0.32).
Both the Central American and Egypt dummy variables make poor choices for
instrumental variables. The Central American dummy variable has a -0.04 correlation
with growth and -0.09 correlation with both aid and aid*policy. Likewise, the Egypt
dummy variable has a -0.12 correlation with growth and 0.02 correlation with both aid
and aid*policy. On the other hand, Arms1 is another relatively good choice of
Resource for conducting the Durbin-Wu-Hausman test in Stata:
http://www.stata.com/support/faqs/statistics/durbin-wu-hausman-test/; November 1999.
19
65
instrumental variables as it has a -0.06 correlation with growth, but a 0.34 correlation
with aid and a 0.32 correlation with aid*policy. Less ideal, Arms1*policy has a low
correlation to growth (-0.07), and an only slightly higher correlation to aid & aid*policy
(both 0.12). Population has a 0.08 correlation to growth, but a -0.24 correlation to aid and
a -0.30 correlation to aid* growth, making it another decent instrumental variable.
However, Pop*policy has a correlation to 0.34 correlation to growth, with a -0.20
correlation to aid and a -0.12 correlation to aid*policy, making it less appealing.
Likewise, Pop2*policy has a 0.29 correlation with growth, with a -0.27 correlation to aid
and a -0.22 correlation to aid*policy. GDP*policy has a 0.24 correlation to growth, but 0.32 with aid and -0.26 with aid*policy. Finally, GDP2*policy has a 0.12 correlation to
growth, but a -0.37 correlation with aid and a -0.33 correlation with aid*policy. Thus, the
GDP2*policy variables is probably the more appropriate instrumental variable of the two.
Not selected in the original works, but also exhibiting low correlation with growth
while having higher correlation with aid and aid*policy, are the GDP variable, the lowincome dummy variable, and the outlier variable. GDP has a 0.02 correlation with
growth, but a -0.39 correlation with aid and a -0.39 correlation with aid* policy. Lowincome has a -0.04 correlation with growth, but a 0.39 correlation with both aid and
aid*policy. Finally, the outlier variable has a 0.05 correlation with growth, but a 0.66
correlation with aid and a 0.67 correlation with aid*policy.
If one has variables to use as instruments, the Durbin-Wu-Hausman augmented test
can be conducted to test for endogeneity, thus verifying the necessity of instrumentation.
To conduct this test, it is necessary to regress the instruments on the assumed endogenous
66
variable, in this case aid and aid* policy. One then predicts the residuals of that
regression and adds those residuals to the right-side variables in a regression of whatever
the original dependent variable was, in this case growth. If the t-value of the residual
coefficient is not statistically significant, that variable does not have endogeneity
problems and does not need to be instrumented.
For each time period, aid and aid*policy were tested separately. In the last time
period, aside from the Easterly-Levine-Roodman and Burnside-Dollar set of instruments,
an alternate set of instruments was selected based on available variables with the lowest
correlation to growth and highest correlation to aid and aid*policy. Those selected were
the French Zone dummy variable, Arms1, Lpop, LGDP2*policy, the outlier variable, and
the low-income variable. The aid or aid*policy coefficients for the last set of estimates
with all time period, for either the original or alternate instrument set, had statistically
significant residuals, indicating an endogenous relationship with growth for both aid and
aid*policy. These results were true for both the GDP deflator and the CPI policy index
estimation results.
As a result of the determination of engdogeneity, the instruments need to be tested for
weakness and the estimation as a whole needs to be tested for overidentification. The
Anderson weak instrument statistic returned a p-value of 0.0000, while the Sargan
statistic for overidetification returned a p-value of 0.0078, indicating both weak
instruments and overidentification of all instruments. Further, the 2SLS estimation
method is best for estimation when data is homoskedastic. It does not do well with both
weak instruments and heteroskedasticity (test results discussed later reveal
67
heteroskedasticity that is mitigated to some degree, but not eliminated by using robust
standard errors).
Therefore an alternate estimation method, the general method of moments (GMM)
and the fixed effects panel (XT) estimation methods were considered. The GMM method
controls better for weak instruments and heteroskedastic data, but the XT method is
designed for panel data.
Ultimately, the XT method was selected and the endogeneity test was run again.
These test results revealed that there is not endogeneity within the model. Thus, the
XT/OLS estimation method, rather than the XT/2SLS method can be used. Aid,
aid*policy and aid2*policy were estimated in all combinations to determine the
estimation that most likely best expresses the relationship between aid-policy-growth.
See Sign Table in Appendix C. It was from these results that the GDP deflator policy
index was selected over the CPI policy index, as the aid and aid-policy interactive terms
are not significant in the estmations, not even when estimated independently when
multicollinearity could not be cancelling out or confusing significance or signs.
Thus, the estimation that best represents the relationship of aid and policy to growth
is the XT/OLS estimation that includes aid, aid*policy and aid2*policy in the GDP
deflator policy index for all time periods with all countries.
68
Chapter 7
FINAL CONCLUSION
Above and beyond anything else, the one undeniable conclusion of this study is that
policy is positive and significantly related to economic growth, with or without
developmental aid. While it does not matter which measure of inflation is considered, as
they are both significant, the policy indicies clearly produce different results and the GDP
deflator seems the more appropriate tool to address GDP related topics. This might be
due to the fact that the CPI only addresses consumer-related inflation, while the GDP
deflator addresses inflation throughout the entire economy. Moreover, in relation to aid
and growth, the other areas of the economy -investment, net exports and government
spending- are the ones that are known or predicted by theory to be influenced by aid. The
two-gap theory states that aid positively influences growth via savings-investment and
import-export gap filling. Additionally, the majority of aid allocations tend to be directly
to governments, often for budgetary support. Aid may trickle into the consumer section
of the economy through one of these other sectors, but it is generally focused on other
areas first.
Further, based on these results, it is possible that aid also has a positive impact on
growth, though this result is not robustly with other estimation methods. Multicollinearily
and endogeneity with the aid and aid-policy variables are two probably reasons for the
lack of clarity, making further articulation of the relationship challenging. Additionally,
the F-test that all u_i=0, that is all sample countries have the same country coefficient,
69
was 2.13, with a p-value of 0.0000. Thus, we reject the null hypothesis that the sample
should be pooled. Further testing is needed to determine the appropriate relationship of
aid to growth, though it should be done with a fixed effects model when testing more
than one country. See the Final Results Table in Appendix D for a more complete
accouinting of estimation of results.
As far as Burnside and Dollar (2000) versus Easterly, Levine, and Roodman (2003),
and who was right or wrong, they were both right and they were both wrong. ELR were
correct that their results, and these, are not robust for the aid-policy-growth relationship.
However, they did not consider the appropriateness of their estimations or what problems
there might be in their data that might be causing that lack of robustness, such as
multicollinearity. Nonetheless, the results here support their finding of the non-robustness
of the effect of aid on growth. BD were accurate in assessing that policy has a positive
role in the growth relationship, and that it might even assist aid in being more effective.
However, it is far from clear that policy is necessary for effective aid. Thus the
recommendation and allocation of developmental aid on that basis over need seems a bit
extreme.
For future research, the expansion of the dataset to include more countries and more
years of data for the countries than are already included would be ideal. As the issue of
endogeneity seems to reoccur between aid and growth, the identification of stronger
instruments seems necessary. Finally, the variables of institutional quality and
ethnolinguistic fractionalization both change over time, but are only expressed as
constant, single year observations in this data set because the long-term country data is
70
not available in usable form. The creation of formal databases for both of these would be
appropriate for such research. Further, there is a database for both assassination and
effective developmental aid data. However, they only extend into the late 1990’s.
Extension and maintenance of these databases would add relevant variable options.
In the end, even with the uncertainty, these results present the developmental aid
question in a more positive light. It is possible that developmental aid contributes
positively to growth. It is certain that policy helps. It is likely that they benefit one
another in contributing to growth. However, neither is necessary for the other.
71
APPENDICES
72
APPENDIX A
DESCRIPTIVE STATISTICS
Variables: ELR Countries, 1966 - 2009
GDP Growth
Aid
Aid*Policy, GDP Deflator
Aid^2*Policy, GDP Deflator
Policy, GDP Deflator
Aid*Policy, CPI
Aid^2*Policy, CPI
Policy, CPI
Budget Surplus
CPI Inflation
Log CPI
GDP Deflator
Log GDP Deflator
Trade Openness
GDP
Initial GDP
Institutional Quality
Lagged Money Supply
Log Population
Lagged Arms Imports
Mean Median
1.660
1.820
0.329
0.167
0.186
0.080
0.016
0.001
6.170
6.310
0.111
0.046
0.009
0.001
3.930
4.070
-0.033 -0.027
0.4412 0.0919
0.174
0.088
0.5385 0.0889
0.178
0.085
57.66
52.275
1628.30 955.14
1317.37 776.23
4.36
4.48
43.21
27.90
16.24
16.17
0.0332 0.0098
Standard Deviation
3.360
0.0427
0.253
0.053
0.792
0.152
0.03
0.769
0.049
3.0472
0.364
3.7889
0.406
32.37
1808.41
1393.17
1.56
255.86
1.42
0.0761
Range
34.900
0.355
2.190
0.772
9.420
1.291
0.428
9.728
0.704
62.5512
4.188
70.249
4.371
200.36
14363.28
6489.99
6.66
6932.06
7.99
0.9360
Minimum
-13.21
-0.001
-0.013
-0.0003
-0.519
-0.088
-0.002
-3.581
-0.45
-0.0362
-0.037
-0.1006
-0.106
5.185
83.08
121.24
1.24
4.16
12.9
-0.0854
Maximum
21.694
0.354
2.179
0.771
8.900
1.203
0.426
6.147
0.254
62.515
4.151
70.148
4.265
205.545
14446.36
6611.23
7.48
6936.22
20.89
0.8501
Count
733
739
580
580
589
570
570
579
633
686
684
707
706
726
719
753
765
731
761
739
CORRELATION VALUES
Growth
Aid
Aid*Polic
Aid^2*Policy
Policy
Budget Surplus
GDP Deflator
Openness
ICRGE
M2
Log Initial GDP
SSA
EAsia
Growth
1.0000
-0.0316
-0.0037
0.1070
0.3387
0.2450
-0.2763
0.1287
0.1183
-0.0598
0.0154
-0.2132
0.2590
Aid
1.0000
0.9909
0.8375
-0.0228
-0.1822
-0.0471
0.1888
-0.1190
-0.0322
-0.3907
0.4449
-0.1872
Aid*Policy Aid^2*Policy
1.0000
0.8394
0.0683
-0.1049
-0.1159
0.2140
-0.0998
-0.0315
-0.3942
0.4473
-0.1853
1.0000
0.0171
-0.1208
-0.0513
0.2308
-0.0701
-0.0103
-0.1323
0.1555
-0.0891
Policy
1.0000
0.6521
-0.8676
0.3646
0.1213
0.0083
-0.0496
0.0357
0.1168
BB
1.0000
-0.2239
0.0407
0.1467
-0.0156
0.0912
0.0341
0.0826
GDP Deflator
1.0000
-0.2463
-0.0610
-0.0227
0.1392
-0.0390
-0.0680
Open
1.0000
0.0383
-0.0027
-0.0290
-0.0527
0.1661
ICRGE
1.0000
-0.0124
0.1615
0.0516
0.1530
M2
1.0000
-0.0066
0.0320
0.0090
Log Initial GDP
1.0000
-0.5935
-0.0494
SSA
1.0000
-0.2271
EAsia
1.0000
73
APPENDIX B
INITIAL RESULT TABLES
Table 1
GDP Deflator Policy, BD Countries and Years
BD gdpd
Variable
bdlgdp
bdethnf
bdssa
1
Coefficient
(t-value)
-1.219 ***
bdm21
bdbb
5/2SLS
5+/2SLS
LOW-INCOME COUNTRY SAMPLE
8/OLS
8+/OLS
8/2SLS
8+/2SLS
7/OLS
*** -1.077 *** -1.095 *** -1.307 *** -1.215 ***
-1.293
*** -1.278 *** -1.300 *** -1.288
** -1.283 ***
( - 4 .4 4 )
( - 4 . 3 1)
( - 4 .6 5)
( - 4 .2 0 )
( - 4 .2 4 )
( - 2 .9 9 )
( - 2 .6 6 )
( - 2 .9 9 )
( - 2 .52 )
-0.858
-0.517
-0.304
-0.422
-0.847
-0.655
-0.657
-0.516
-0.627
-0.798
-0.613
( - 1. 15 )
( - 0 .70 )
( - 0 . 4 1)
( - 0 .59 )
( - 0 .9 5)
( - 0 .79 )
( - 0 .72 )
( - 0 .55)
( - 0 .70 )
( - 0 .73 )
( - 0 .6 3 )
-2.781 ***
2.043 ***
( 3 .4 8 ) .
bdicrge
-1.139
FULL SAMPLE
5/OLS
5+/OLS
( - 5 . 19 )
( - 5 . 12 )
bdeasia
4/OLS
0.708 ***
-3.330
*** -3.286 *** -3.487 *** -2.204 ** -3.078 ***
( - 4 .4 2 )
( - 4 . 3 1)
1.971
*** 2.038 ***
( 3 .4 8 ) .
0.771
( - 5.0 6 )
( 3 .54 ) .
( - 2 . 14 )
( - 3 .52 )
( 3 .3 0 ) .
*** -3.521 *** -3.442 *** -2.985 *** -3.473 ***
( - 4 .0 5)
2.011 *** 1.899 *** 1.950 ***
( 3 .55) .
-3.386
( 3 .4 8 ) .
2.392
( - 4 . 12 )
0.848
( - 4 .4 7)
( - 2 .74 )
( - 4 .0 0 )
*** 2.554 *** 2.432 *** 2.540 *** 2.446 ***
( 3 .4 0 ) .
*** 0.808 *** 0.783 *** 0.718 *** 0.744 ***
( - 2 .6 2 )
( 3 . 5 1) .
( 3 .50 ) .
( 3 .3 5) .
( 3 .53 ) .
*** 0.856 *** 0.845 *** 0.814 *** 0.849 ***
( 4 . 4 1) .
( 4 .52 ) .
( 4 .70 ) .
( 4 .73 ) .
( 3 .75) .
( 4 . 14 ) .
( 3 .8 2 ) .
( 3 .59 ) .
( 3 .75) .
( 3 .3 0 ) .
0.004
-0.002
-0.007
-0.004
0.006
-0.001
0.014
0.011
0.013
0.016
( 3 .6 2 ) .
0.013
( 0 .2 4 ) .
( - 1. 10 )
( - 0 .4 0 )
( - 0 .2 3 )
( 0 .2 6 ) .
( - 0 .0 5)
( 0 .70 ) .
( 0 .54 ) .
( 0 .6 8 ) .
( 0 .70 ) .
( 0 .6 7) .
64.709
50.659
75.004
21.858
96.395
51.734
-1.816
56.913
-90.973
49.671
( 1. 2 7 ) .
( 0 .6 0 ) .
( 1. 5 1) .
( 0 . 17 ) .
( 1. 0 2 ) .
( 0 .8 9 ) .
( - 0 .0 2 )
( 1. 0 5 ) .
( - 0 .54 )
( 0 .53 ) .
13.828 ***
( 4 .4 8 ) .
bdinfl2
-0.359
( - 1. 0 4 )
bdsacw
0.006
( 0 . 9 1) .
bdaid
bdaidpolicy2
bdaid2policy2
bdpolicy2_b1
-7.419
-4.507
-7.292
-4.919
-11.000
-5.195
2.383
-5.276
11.338
-4.310
( - 1. 19 )
( - 0 .4 2 )
( - 1. 16 )
( - 0 .3 3 )
( - 0 .9 8 )
( - 0 .76 )
( 0 . 19 ) .
( - 0 .77)
( 0 .57) .
( - 0 .3 8 )
9.442
3.768
( 0 .4 7) .
( 0 . 17 ) .
1.302
( 4 .4 4 ) .
Period 3
Period 4
Period 5
Period 6
Period 7
Constant
*** 1.254 *** 1.342 ***
( 3 .50 ) .
1.12
**
( 2 .3 4 ) .
1.421 ***
( 3 .3 5) .
1.156
*** 0.968
( 3 .0 5) .
*
( 1. 9 4 ) .
1.181
*** 0.576
( 3 . 16 ) .
1.144
( 0 .79 ) .
-0.13
-0.25
-0.43
-0.23
-0.46
-0.17
-0.12
-0.23
-0.12
-0.23
-0.12
( - 0 .4 6 )
( - 0 .8 0 )
( - 0 .4 2 )
( - 0 .8 4 )
( - 0 . 3 1)
( - 0 . 16 )
( - 0 . 3 1)
( - 0 . 16 )
( - 0 . 3 1)
( - 0 . 17 )
-1.080
-1.21
-1.06
-1.21
-0.95
-1.12
-1.51
-1.31
-1.51
-1.18
-1.53
( - 1. 6 2 )
( - 1. 8 2 )
( - 1. 5 4 )
( - 1. 8 2 )
( - 1. 3 5 )
( - 1. 6 9 )
( - 1. 6 8 )
( - 1. 3 8 )
( - 1. 6 9 )
( - 1. 2 4 )
( - 1. 7 4 )
-3.21
-3.24
-5.16
-3.22
-3.1
-3.13
-3.05
-2.98
-3.04
-2.96
-3.06
( - 5.2 4 )
( - 5 . 2 1)
( - 5.0 8 )
( - 5 . 2 1)
( - 4 .9 0 )
( - 5 . 11)
( - 3 .70 )
( - 3 .6 0 )
( - 3 .72 )
( - 3 .73 )
( - 3 .9 3 )
-1.78
-1.75
-1.76
-1.73
-1.84
-1.69
-1.87
-1.97
-1.87
-1.99
-1.88
( - 2 .9 5)
( - 3 .0 0 )
( - 3 .0 2 )
( - 2 .9 6 )
( - 2 .9 9 )
( - 2 .9 0 )
( - 2 .4 2 )
( - 2 .54 )
( - 2 .4 2 )
( - 2 .53 )
( - 2 .4 8 )
-2.31
-2.45
-2.23
-2.43
-2.22
-2.38
-3.44
-3.35
-3.44
-3.24
-3.45
( - 3 .4 6 )
( - 3 .76 )
( - 3 .2 5)
( - 3 .76 )
( - 3 .2 8 )
( - 3 .72 )
( - 4 .0 4 )
( - 3 .6 4 )
( - 4 .0 4 )
( - 3 .53 )
( - 4 . 11)
8.84
-3.41
-3.73
-4.21
-0.32
-3.78
-1.84
-0.54
-2.06
3.19
-1.89
( 4 .4 2 ) .
( - 1. 0 2 )
( - 1. 0 9 )
( - 1. 4 3 )
( - 0 .0 6 )
( - 0 .8 4 )
( - 0 .4 0 )
( - 0 . 10 )
( - 0 .4 8 )
( 0 .4 3 ) .
( - 0 .3 5)
^^^
^^^
277
0.4126
17.70
1.73
277
0.4283
16.40
1.70
258
0.4263
16.58
1.73
^^^
^^^
277
0.4275
17.38
1.73
258
0.4083
18.92
1.73
^^^
^^^
^^^
^^^
^^^
^^^
^^^
277
0.4233
19.70
1.73
192
0.4455
11.08
1.72
175
0.4472
11.02
1.75
192
0.4453
11.87
1.75
175
0.4402
12.24
1.75
**
( 2 .3 4 ) .
( - 0 .2 2 )
Only Lo w inc o me
Out lie rs Omit t e d
2 S LS
Observations
R^2
F-Stat
F-Crit at 95%
( 4 .57) .
^^^
^^^
192
0.4453
13.36
1.75
74
Table 2
CPI Policy, BD Countries and Years
BD cpi
Variable
bdlgdp
bdethnf
bdssa
1
Coefficient
(t-value)
-1.007 ***
bdm21
bdbb
5+/2SLS
LOW-INCOME COUNTRY SAMPLE
8/OLS
8+/OLS
8/2SLS
8+/2SLS
7/OLS
-0.962 *** -0.911 *** -0.846 *** -1.129 *** -0.951 ** -1.285 *** -1.312 *** -1.290 *** -1.380 *** -1.291 ***
( - 2 .9 4 )
( - 2 .79 )
- 2 .6 5
( - 2 .6 6 )
( - 2 .3 0 )
( - 2 .9 6 )
( - 2 .72 )
( - 2 .9 9 )
( - 2 .6 5)
-0.831
-0.569
-0.516
-0.386
-1.000
-0.738
-0.715
-0.929
-0.630
-1.400
-0.648
( - 1. 0 7 )
( - 0 .74 )
( - 0 .6 5)
( - 0 . 5 1)
( - 1. 0 4 )
( - 0 .8 0 )
( - 0 .75)
( - 0 . 9 1)
( - 0 .6 7)
( - 1. 11)
( - 0 .6 4 )
-2.446 ***
2.145 ***
( 3 .9 6 ) .
bdicrge
5/2SLS
( - 3 .2 6 )
( - 3 .54 )
bdeasia
FULL SAMPLE
5/OLS
5+/OLS
4/OLS
0.767 ***
-2.713
*** -2.679 *** -3.122 *** -1.832
( - 3 . 11)
2.098
( - 3 . 0 1)
*** 2.239 *** 2.164 ***
( 3 . 9 1) .
0.779
( - 3 .8 7)
( 4 .0 9 ) .
( 3 .9 9 ) .
*
( - 1. 7 8 )
( - 2 .78 )
-2.602 *** -3.191 *** -3.268 *** -3.414 *** -2.609 ** -3.398 ***
( - 2 .6 7)
( - 4 .0 3 )
( - 4 .0 6 )
( - 4 .6 6 )
( - 2 .4 2 )
( - 4 .0 4 )
2.151 *** 2.191 *** 2.183 *** 2.453 *** 2.280 *** 2.434 *** 2.295 ***
( 3 . 8 1) .
( 3 .8 8 ) .
( 3 .58 ) .
( 3 .8 2 ) .
( 3 .74 ) .
( 3 .52 ) .
( 3 .70 ) .
*** 0.858 *** 0.824 *** 0.794 *** 0.775 *** 0.819 *** 0.887 *** 0.827 *** 0.864 *** 0.829 ***
( 4 .72 ) .
( 4 .6 4 ) .
( 5 . 11) .
( 5.0 2 ) .
( 4 .4 9 ) .
( 4 .3 9 ) .
( 3 .73 ) .
( 3 . 9 1) .
( 3 .78 ) .
( 3 .8 2 ) .
-0.003
-0.006
-0.019
-0.011
-0.011
-0.004
0.009
-0.005
0.008
-0.004
( 3 .74 ) .
0.008
( - 0 . 17 )
( - 0 .3 9 )
( - 1. 2 5 )
( - 0 .75)
( - 0 .6 6 )
( - 0 .2 4 )
( 0 .50 ) .
( - 0 .2 8 )
( 0 .4 5) .
( - 0 . 19 )
( 0 .4 5) .
10.022 ***
( 2 .74 ) .
bdinfl1
-0.485
( - 1. 13 )
bdsacw
0.005
( 0 . 5 1) .
bdaid
bdaidpolicy1
bdaid2policy1
bdpolicy1
10.067
-69.719
53.888
-127.732
-24.195
62.731
-88.108
101.022
-223.511
90.398
( 0 . 15 ) .
( - 0 .6 2 )
( 0 .8 9 ) .
( - 0 .8 3 )
( - 0 . 18 )
( 0 .8 9 ) .
( - 0 .6 0 )
- 1. 6 4
( - 0 .8 6 )
( 0 . 7 1) .
-3.867
10.720
-5.444
14.671
4.099
-10.071
14.352
-12.455
30.444
-10.956
( - 0 .4 2 )
- 0 .6 3
( - 0 .6 0 )
( 0 .6 5) .
( 0 . 2 1) .
( - 1. 0 6 )
( 0 .6 6 ) .
( - 1. 3 2 )
( 0 .8 4 ) .
( - 0 .58 )
35.643
23.937
( 1. 4 9 ) .
( 0 .9 4 ) .
1.130
( 2 .3 8 ) .
Period 3
Period 4
Period 5
Period 6
Period 7
Constant
**
0.831
( 2 .8 5) .
0.895
( 0 .79 ) .
1.668 *** 0.931
( 1. 2 3 ) .
( 3 .0 7) .
1.868 *** 0.192
( 1. 2 3 ) .
( 3 .6 7) .
1.802
( 0 . 15 ) .
-0.24
-0.40
-0.41
-0.32
-0.45
-0.26
-0.26
-0.18
-0.23
-0.18
-0.23
( - 0 .59 )
( - 0 . 6 1)
( - 0 .4 7)
( - 0 .6 6 )
( - 0 .3 8 )
( - 0 .3 3 )
( - 0 .2 3 )
( - 0 .2 9 )
( - 0 .2 2 )
( - 0 .2 8 )
-1.50
-1.61
-1.38
-1.60
-1.28
-1.53
-1.23
-0.97
-1.25
-0.78
-1.25
( - 2 .0 2 )
( - 2 . 18 )
( - 1. 8 3 )
( - 2 . 16 )
( - 1. 6 7 )
( - 2 .0 5)
( - 1. 2 8 )
( - 0 .9 8 )
( - 1. 3 1)
( - 0 .78 )
( - 1. 3 3 )
-3.41
-3.45
-3.44
-3.39
-3.47
-3.45
-2.60
-2.64
-2.57
-2.65
-2.58
( - 5 . 11)
( - 5.0 8 )
( - 5.0 3 )
( - 5.0 4 )
( - 4 .9 7)
( - 4 .9 0 )
( - 3 .0 7)
( - 3 .0 0 )
( - 3 .0 4 )
( - 3 . 10 )
( - 3 . 14 )
-1.90
-1.96
-1.79
-1.86
-1.82
-1.92
-1.56
-1.40
-1.53
-1.31
-1.54
( - 2 .9 3 )
( - 3 .0 8 )
( - 2 .8 3 )
( - 2 .9 3 )
( - 2 .8 4 )
( - 2 .9 3 )
( - 1. 9 6 )
( - 1. 7 3 )
( - 1. 9 1)
( - 1. 6 3 )
( - 1. 9 5 )
-2.35
-2.53
-2.17
-2.45
-2.13
-2.36
-3.11
-2.88
-3.10
-2.69
-3.09
( - 3 .3 3 )
( - 3 .6 7)
( - 3 .0 4 )
( - 3 .56 )
( - 3 .0 2 )
( - 3 . 4 1)
( - 3 .54 )
( - 3 .0 2 )
( - 3 . 5 1)
( - 2 .76 )
( - 3 .4 7)
7.32
-1.14
0.25
-3.67
4.40
0.26
-3.61
1.60
-5.31
7.83
-4.86
( 2 .9 7) .
( - 0 .3 0 )
( 0 .0 7) .
( - 1. 13 )
( 0 .79 ) .
( 0 .0 5) .
( - 0 .70 )
( 0 .2 5) .
( - 1. 15 )
( 0 .73 ) .
( - 0 .77)
^^^
^^^
272
0.4635
16.21
1.73
272
0.383
15.60
1.70
252
0.3862
15.73
1.73
^^^
^^^
272
0.3776
16.07
1.73
252
0.3736
17.43
1.73
^^^
^^^
^^^
^^^
^^^
^^^
^^^
272
0.3673
17.36
1.73
185
0.4440
12.95
1.72
167
0.4525
12.76
1.75
185
0.4413
13.48
1.75
167
0.4413
13.54
1.75
**
( 2 .2 9 ) .
( - 0 .3 5)
Only Lo w inc o me
Out lie rs Omit t e d
2 S LS
Observations
R^2
F-Stat
F-Crit at 95%
1.304 *** 0.550
( 1. 5 2 ) .
^^^
^^^
185
0.4412
15.15
1.75
75
Table 3
GDP Deflator Pilicy, ELR Countries, BD Years
FULL SAMPLE
5/OLS
5+/OLS
P o lic y 2
Variable
elrlgdp
elrethnf
elrssa
1
Coefficient
(t-value)
-0.682 ***
( - 3 . 19 )
-1.344
-1.484
( - 1. 5 7 )
( - 1. 6 7 )
-1.621 ***
2.38
0.315
***
**
2.639
***
0.395
-0.001 ***
( - 2 .54 )
( - 3 .2 5)
( - 2 .52 )
-0.973
-1.327
-1.161
-1.571
( - 1. 11)
( - 1. 5 0 )
( - 1. 2 7 )
( - 1. 6 9 )
-2.209
2.56
0.392
-0.001 *** -0.001
LOW-INCOME COUNTRY SAMPLE
8/OLS
8+/OLS
8/2SLS
8+/2SLS
7/OLS
*
*
-0.953 ** -0.523
-0.936 ** -0.519
( - 1. 8 5 )
( - 2 .50 )
( - 1. 4 0 )
( - 2 .3 2 )
-1.265
-0.859
-1.118
-0.855
( - 1. 3 9 )
-1.024
( - 1. 2 1)
( - 0 .8 5)
( - 1. 0 9 )
( - 0 .8 2 )
( - 0 .9 9
*** -1.97 *** -1.941 *** -1.742 *** -1.467 ** -1.937 *** -1.800 *** -1.953
( - 3 .0 7)
***
( - 2 .59 )
( - 2 .6 5)
( - 2 .0 7)
2.86 *** 2.494 *** 2.776 ***
( 5.4 5) .
( 4 .8 9 ) .
( 5.2 4 ) .
*** 0.382 ** 0.385 *** 0.363
( 2 .8 0 ) .
( - 7 . 10 )
5+/2SLS
*** -0.609 ** -0.946 *** -0.689 ** -0.719
( 5.0 6 ) .
***
5/2SLS
( - 3 .50 )
( - 3 .4 6 )
( 2 .70 ) .
( - 8 .73 )
elrbb
**
( 5 . 2 1) .
( 2 .2 8 ) .
elrm21
-1.631
*
( - 2 .4 0 )
( 4 .53 ) .
elricrge
-0.791 *** -0.849
( - 2 .9 0 )
( - 2 .9 6 )
elreasia
4/OLS
( 2 .54 ) .
( 2 .74 ) .
2.98
( - 3 .0 3 )
( 2 . 4 1) .
0.471
-1.900 ***
( - 2 .4 8 )
( - 2 .77)
*** 2.956 *** 3.252 *** 2.982 *** 3.305 ***
( 5.4 0 ) .
**
( - 2 .6 7)
( 5.3 8 ) .
**
( 2 .4 9 ) .
( 5.6 5) .
0.487 *** 0.446
( 2 .8 8 ) .
( 5 . 17 ) .
**
( 2 .2 7) .
( 5.3 5) .
0.482 *** 0.450
( 2 .8 6 ) .
**
( 2 .3 0 ) .
*** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 ***
( - 6 .55)
( - 7.0 6 )
( - 5.54 )
( - 6 .3 6 )
( - 6 .2 6 )
( - 5.9 4 )
( - 6 .2 4 )
( - 4 .6 7)
( - 5.0 5)
10.179 ***
( 2 .9 8 ) .
klinfl
-0.876 **
( - 2 .0 6 )
elrsacw
0.016
**
( 2 . 12 ) .
elraid
elraidpolicy2
elraid2policy2
1.700
3.012
2.026
-8.333
-2.564
5.200
5.991
1.822
1.572
8.146
( 0 . 18 . )
( 0 .3 7) .
( 0 . 18 ) .
( - 0 .6 0 )
( - 0 . 19 )
( 0 .3 8 ) .
( 0 .53 ) .
( 0 . 11) .
( 0 . 10 ) .
( 0 .4 8 ) .
-1.120
-0.211
1.506
0.310
1.255
-1.787
-1.186
1.527
-0.440
1.013
( - 0 .76 )
( - 0 .2 0 )
( 1. 11) .
( 0 . 2 1) .
( 0 .8 2 ) .
( - 0 .8 7)
( - 0 . 8 1)
( 0 .74 ) .
( - 0 .2 5)
( 0 .53 ) .
10.992
**
11.571
( 2 . 4 1) .
elrpolicy2
0.222
( 3 . 12 ) .
Period 3
Period 4
Period 5
Period 6
Period 7
Constant
***
0.206
*** 0.158 **
( 3 . 19 ) .
( 2 .2 6 ) .
0.186 *** 0.168
( 2 .70 ) .
**
( 2 .2 2 ) .
0.238
**
( 2 .2 3 ) .
0.229
**
( 2 .4 3 ) .
0.140
0.200 **
( 1. 4 2 ) .
( 2 .0 8 ) .
0.160
( 1. 6 0 ) .
0.53
-0.81
0.26
-0.82
-0.22
-0.72
-0.90
-0.24
-0.92
-0.25
-0.97
( - 0 .78 )
( - 1. 2 1)
( - 0 .4 3 )
( - 1. 2 3 )
( - 0 .3 6 )
( - 1. 0 8 )
( - 1. 0 1)
( - 0 . 3 1)
( - 1. 0 4 )
( - 0 .3 3 )
( - 1. 12 )
-1.69
-2.1
-1.61
-2.14
-1.52
-2.01
-2.08
-1.49
-2.15
-1.51
-2.21
( - 2 . 18 )
( - 2 .70 )
( - 2 .2 6 )
( - 2 .75)
( - 2 . 13 )
( - 2 .6 0 )
( - 2 .0 6 )
( - 1. 6 6 )
( - 2 . 14 )
( - 1. 7 4 )
( - 2 .2 8 )
-3.24
-3.59
-2.96
-3.71
-2.88
-3.61
-3.11
-2.22
-3.32
-2.25
-3.36
( - 4 . 6 1)
( - 5.0 9 )
( - 4 .9 3 )
( - 5.3 2 )
( - 5 . 0 1)
( - 5.3 6 )
( - 3 .3 0 )
( - 2 .8 4 )
( - 3 .59 )
( - 3 .0 3 )
( - 3 .8 3 )
-2.09
-2.18
-1.55
-2.29
-1.48
-2.22
-1.99
-1.13
-2.14
-1.16
-2.18
( - 3 .0 6 )
( - 3 . 17 )
( - 2 .56 )
( - 3 .3 3 )
( - 2 .53 )
( - 3 .3 0 )
( - 2 . 19 )
( - 1. 4 6 )
( - 2 .3 8 )
( - 1. 5 8 )
( - 2 .53 )
-2.09
-2.02
-1.47
-2.13
-1.36
-2.04
-2.83
-2.12
-2.96
-2.12
-3.03
( - 3 .0 2 )
( - 2 .8 2 )
( - 2 .3 6 )
( - 3 .0 0 )
( - 2 .2 0 )
( - 2 .9 4 )
( - 2 .9 8 )
( - 2 .6 0 )
( - 3 . 16 )
( - 2 . 6 1)
( - 3 .3 6 )
6.98
6.24
6.01
5.21
7.00
5.94
5.14
5.82
4.30
5.93
4.03
( 3 .8 5) .
( 2 .9 5) .
( 2 .9 2 ) .
( 2 .55) .
( 2 .77) .
( 2 .4 5) .
( 1. 6 5 ) .
( 1. 9 1) .
( 1. 4 2 ) .
( 1. 8 1) .
( 1. 2 9 ) .
Only Lo w inc o me
Out lie rs Omit t e d
2 S LS
Observations
R^2
F-Stat
F-Crit at 95%
**
( 2 .3 4 ) .
^^^
^^^
324
0.3198
70.08
1.72
318
0.3298
55.78
1.70
289
0.3271
64.95
1.73
^^^
^^^
318
0.3189
50.72
1.72
289
0.3236
70.96
1.73
^^^
^^^
^^^
^^^
^^^
^^^
^^^
318
0.3143
62.90
1.72
225
0.3508
56.69
1.71
198
0.3709
59.91
1.74
225
0.3362
52.60
1.74
198
0.3705
59.42
1.74
^^^
^^^
225
0.3354
54.08
1.74
76
Table 4
CPI Policy, ELR Countries, BD Years
Policy 1
Variable
elrlgdp
elrethnf
elrssa
1
Coefficient
(t-value)
-0.456
-0.587
( - 1. 5 3 )
( - 1. 8 2 )
-0.298
-0.775 * -0.409
( - 0 .9 4 )
( - 1. 9 5 )
-1.105
( - 1. 14 )
*** -0.952 ** -0.787 ** -0.962 ** -0.754
( - 2 .9 5)
( - 2 .3 8 )
( - 2 .0 9 )
( - 2 .2 5)
-0.506
-0.614
-1.009
-0.971
-0.970
-0.460
-0.599
-0.975
-0.786
( - 0 . 6 1)
( - 0 .73 )
( - 1. 0 9 )
( - 1. 0 7 )
( - 1. 0 3 )
( - 0 .4 9 )
( - 0 .6 5)
( - 0 .9 2 )
( - 0 . 8 1)
-1.685
** -0.392
-1.250
( - 2 .2 7)
( - 0 .4 7)
( - 1. 7 2 )
-1.379
**
2.666 ***
0.255
-1.317
( - 1. 6 6 )
2.533
**
0.298
( 4 .52 ) .
**
( 2 .2 4 ) .
-0.001
*
*
( 5 . 14 ) .
0.422 *** 0.275
( 3 . 19 ) .
( 1. 9 9 ) .
( 4 .3 9 ) .
**
( 5.0 6 ) .
( - 5.6 8 )
( - 5.6 5)
( 1. 5 4 ) .
( - 6 .59 )
2.642
0.318
( - 6 .2 9 )
-0.001
-2.131 *** -2.209 *** -1.217
( - 3 .2 3 )
( - 3 .57)
-1.948 ***
( - 1. 3 7 )
( - 2 .6 6 )
*** 2.613 *** 3.091 *** 2.636 *** 3.202 ***
( 4 . 6 1) .
**
( 2 .0 3 ) .
*** -0.001 *** -0.001 *** -0.001 *** -0.001 ***
( - 6 .2 9 )
**
( 4 .6 8 ) .
0.364 *** 0.216
( 2 .70 ) .
-1.651
( - 2 .53 )
*** 2.406 *** 2.840 *** 2.419 *** 2.920 ***
( 4 .8 2 ) .
-0.001 ***
8.110
-1.187
( - 1. 5 3 )
*
( - 1. 9 5 )
-0.938
( - 6 .4 7)
elrbb
*
LOW-INCOME COUNTRY SAMPLE
8/OLS
8+/OLS
8/2SLS
8+/2SLS
7/OLS
( - 1. 12 )
( 2 . 0 1) .
elrm21
-0.536
( - 1. 6 6 )
5+/2SLS
-1.060
( 5 . 0 1) .
elricrge
*
5/2SLS
( - 1. 3 1)
( - 2 .2 2 )
elreasia
FULL SAMPLE
5/OLS
5+/OLS
4/OLS
( 5 . 16 ) .
( 4 .53 ) .
0.473 *** 0.267
( 3 .0 3 ) .
( 5.0 0 ) .
0.462 *** 0.257
( 1. 5 8 ) .
( 2 .9 3 ) .
( 1. 5 0 ) .
*** -0.001 *** -0.001 *** -0.001 *** -0.001 ***
( - 6 .52 )
( - 5.77)
( - 5.76 )
( - 5.3 3 )
( - 4 .75)
58.241
36.017
44.059
-105.42
-24.9
( 1. 5 3 ) .
( 0 .8 8 ) .
( 1. 2 1) .
( - 1. 0 7 )
( - 0 . 3 1)
*
( 1. 6 7 ) .
elrinfl
-1.478 ***
( - 2 .6 7)
elrsacw
0.006
( 0 .72 ) .
elraid
elraidpolicy1
elraid2policy1
6.271
-18.448
-1.283
( 0 .2 5) .
( - 0 .58 )
( - 0 .0 5)
-117.64 ** -76.65
( - 2 .2 6 )
-4.746
2.045
2.288
16.452
14.794
( - 0 .9 5)
( 0 .3 3 ) .
( 0 .4 7) .
( 1. 6 1) .
( 1. 7 7 ) .
23.701
*
( - 1. 6 9 )
*
1.046
( 3 .6 3 ) .
Period 3
Period 4
Period 5
Period 6
Period 7
Constant
***
-7.042
-5.82
15.96
6.531
( - 0 .8 7)
( - 0 .8 3 )
( 0 .8 8 ) .
( 0 .4 5) .
26.885 ***
( 3 .9 0 ) .
*** 0.944 *** 0.903 *** 0.542 ** 0.543
( 3 . 4 1) .
( 3 .2 5) .
( 2 . 19 ) .
**
( 2 .2 0 ) .
1.674
*** 1.453
( 2 .6 4 ) .
**
( 2 .4 0 ) .
1.405
**
( 2 .4 7) .
0.467
0.850
( 0 .57) .
( 1. 15 ) .
0.21
0.05
-0.06
0.03
-0.01
0.16
0.07
0.04
-0.02
0.07
-0.01
( 0 .3 0 ) .
( - 0 .0 8 )
( - 0 .0 8 )
( 0 .0 5) .
( - 0 .0 2 )
( 0 .2 4 ) .
( 0 .0 9 ) .
( 0 .0 6 ) .
( - 0 .0 2 )
( 0 .0 9 ) .
( - 0 . 0 1)
-1.380
-1.61
-1.58
-1.66
-1.45
-1.53
-1.01
-0.96
-1.16
-0.83
-1.17
( - 1. 9 4 )
( - 2 .2 3 )
( - 2 . 17 )
( - 2 .2 8 )
( - 1. 9 4 )
( - 2 .0 5)
( - 1. 16 )
( - 1. 0 8 )
( - 1. 2 9 )
( - 0 .9 5)
( - 1. 3 2 )
-2.81
-2.79
-2.76
-2.99
-2.66
-2.95
-1.7
-1.75
-2.09
-1.75
-2.18
( - 4 . 7 1)
( - 4 .58 )
( - 4 .4 6 )
( - 4 .8 5)
( - 4 .2 9 )
( - 4 .8 5)
( - 2 .2 4 )
( - 2 .2 7)
( - 2 .6 7)
( - 2 .3 2 )
( - 2 .8 6 )
-1.65
-1.61
-1.62
-1.81
-1.45
-1.79
-0.99
-1.14
-1.37
-0.89
-1.39
( - 2 .77)
( - 2 .6 7)
( - 2 .6 8 )
( - 2 . 9 1)
( - 2 .3 4 )
( - 2 .9 2 )
( - 1. 3 6 )
( - 1. 5 3 )
( - 1. 7 6 )
( - 1. 2 4 )
( - 1. 8 4 )
-1.51
-1.47
-1.39
-1.63
-1.01
-1.45
-1.80
-1.87
-2.14
-1.34
-2.02
( - 2 .4 2 )
( - 2 .3 6 )
( - 2 .2 0 )
( - 2 .58 )
( - 1. 5 3 )
( - 2 .2 8 )
( - 2 .3 4 )
( - 2 .3 7)
( - 2 .6 7)
( - 1. 6 6 )
( - 2 .4 7)
5.53
0.60
0.00
-0.97
4.39
2.14
0.23
-0.57
-0.80
5.20
2.10
( 2 .58 ) .
( 0 .2 2 ) .
( 0 .0 0 ) .
( - 0 .3 7)
( 1. 4 5 ) .
( 0 .76 ) .
( 0 .0 5) .
( - 0 . 14 )
( - 0 .2 0 )
( 1. 0 1) .
( 0 .4 4 ) .
Only Lo w inc o me
Out lie rs Omit t e d
2 S LS
Observations
R^2
F-Stat
F-Crit at 95%
*
( - 1. 9 6 )
( 3 .58 ) .
elrpolicy1
-15.055
^^^
^^^
311
0.2928
71.08
1.73
305
0.327
72.55
1.70
291
0.3192
65.80
1.73
^^^
^^^
305
0.3043
64.17
1.73
291
0.2868
76.87
1.73
^^^
^^^
^^^
^^^
^^^
^^^
^^^
305
0.288
66.28
1.73
208
0.4180
55.27
1.72
196
0.4074
64.79
1.75
208
0.3772
46.70
1.75
196
0.3595
66.85
1.75
^^^
^^^
208
0.3675
48.46
1.75
77
Table 5
GDP Deflator Policy, ELR Countries and Years
FULL SAMPLE
5/2SLS
5+/OLS
5/OLS
Policy4
4/OLS
1
Variable
LOW-INCOME COUNTRY SAMPLE
8+/2SLS
8/2SLS
8+/OLS
8/OLS
7/OLS
5+/2SLS
Coefficient
(t-value)
elrlgdp
elrethnf
elrssa
0.030
0.019
-0.021
0.046
-0.020
0.062
-0.004
-0.053
0.048
-0.010
0.090
( 0 . 16 ) .
( 0 . 11) .
( - 0 . 12 )
( 0 .2 5) .
( - 0 . 11)
( 0 .3 2 ) .
( - 0 .0 2 )
( - 0 .2 5)
( 0 .2 3 ) .
( - 0 .0 5)
( 0 .4 2 ) .
-0.427
-0.131
0.271
-0.119
0.116
-0.328
0.462
0.818
0.319
0.506
0.068
( - 0 .55)
( - 0 . 18 )
( 0 .3 8 ) .
( - 0 . 16 )
( 0 . 15 ) .
( - 0 . 4 1)
( 0 .55) .
( 1. 0 0 ) .
( 0 .3 8 ) .
( 0 .57) .
( 0 .0 7) .
-0.947
*
elreasia
2.491
***
elrm21
elrbb
*
2.583
0.209
0.258
( 1. 6 0 ) .
( 1. 9 8 ) .
-2.086
***
***
2.568
***
0.336
**
2.748
**
***
0.255
*
2.686
***
*
0.269
-0.912
-1.096
-1.912
( - 1. 17 )
( - 1. 6 4 )
( - 2 .9 6 )
2.738
*
*
( 1. 9 6 ) .
2.655
***
0.187
0.309
( 1. 3 8 ) .
( 1. 7 2 ) .
2.694
***
*
0.403
-1.353
**
***
2.950
***
**
**
3.168
0.289
0.385
( 1. 5 7 ) .
( 2 .0 8 ) .
-1.097
( - 1. 4 5 )
***
2.976
**
0.285
( 1. 5 5 ) .
0.003
-0.005
-0.009
-0.004
-0.006
0.001
0.008
0.003
0.008
-0.003
0.007
( 0 . 2 1) .
( - 0 .52 )
( - 0 .9 4 )
( - 0 .4 3 )
( - 0 .54 )
( 0 .0 6 ) .
( 0 .6 3 ) .
( 0 .2 3 ) .
( 0 .59 ) .
( - 0 . 19 )
( 0 .50 ) .
10.277
***
( 4 .75) .
( 4 .8 9 ) .
( 5 . 18 ) .
( 2 .2 2 ) .
-2.234
( - 2 .6 8 )
( - 2 .0 7)
( 4 .79 ) .
( 4 .73 ) .
( 1. 6 7 ) .
( 4 .6 6 ) .
( 1. 9 2 ) .
( 2 .56 ) .
-1.592
( - 1. 8 1)
( 5.0 9 ) .
( 4 . 8 1) .
**
-1.611
( - 2 .58 )
( - 3 .3 5)
( 4 .8 9 ) .
( 4 .6 6 ) .
elricrge
-1.272
( - 1. 9 3 )
( - 1. 7 8 )
**
( 2 .3 9 ) .
klinfl
-0.772
**
( - 2 .3 2 )
elrsacw
0.012
( 1. 6 0 ) .
elraid
elraidpolicy4
20.642
19.05
11.550
-37.74
-36.31
14.895
12.88
1.335
-49.42
-31.600
( 1. 0 2 ) .
( 0 .8 3 ) .
( 0 .58 ) .
( - 1. 6 1)
( - 1. 6 4 )
( 0 .6 0 ) .
( 0 .4 4 ) .
( 0 .0 5) .
( - 1. 2 4 )
( - 0 .9 7)
-8.507
-0.412
0.591
18.344
( - 1. 14 )
( - 0 .0 5)
( 0 . 10 ) .
( 1. 8 0 ) .
25.452
elraid2policy4
*
13.976
*
( 1. 8 8 ) .
1.1430
( 4 . 2 1) .
Period 3
Period 4
Period 5
Period 6
Period 7
Period 8
Constant
3.323
29.249
( 0 .4 2 ) .
( 1. 8 0 ) .
*
13.821
( 1. 2 8 ) .
**
( 2 . 18 ) .
***
1.045
***
0.918
( 3 .6 5) .
( 3 .8 3 ) .
***
0.495
0.506
1.116
( 1. 5 7 ) .
( 1. 6 2 ) .
( 2 . 6 1) .
**
0.980
**
0.740
*
( 1. 9 3 ) .
( 2 .2 9 ) .
-0.114
0.270
( - 0 . 18 )
( 0 .4 8 ) .
-0.57
-0.80
-0.73
-0.78
-0.64
-0.63
-0.62
-0.55
-0.63
-0.61
-0.60
( - 0 .74 )
( - 1. 0 8 )
( - 1. 0 2 )
( - 1. 0 5 )
( - 0 .8 6 )
( - 0 .8 3 )
( - 0 .6 5)
( - 0 .6 0 )
( - 0 .6 6 )
( - 0 .6 6 )
( - 0 .6 3 )
-1.91
-2.13
-2.00
-2.13
-1.92
-1.98
-2.07
-1.91
-2.08
-2.03
-2.05
( - 2 . 17 )
( - 2 .4 4 )
( - 2 .2 8 )
( - 2 .4 2 )
( - 2 . 19 )
( - 2 .2 5)
( - 1. 8 7 )
( - 1. 7 3 )
( - 1. 8 9 )
( - 1. 8 7 )
( - 1. 9 2 )
-3.35
-3.37
-3.18
-3.47
-3.26
-3.47
-3.01
-2.75
-3.18
-2.99
-3.226
( - 4 .2 2 )
( - 4 .2 5)
( - 4 .0 0 )
( - 4 .3 7)
( - 4 .0 6 )
( - 4 .3 8 )
( - 2 .9 3 )
( - 2 .72 )
( - 3 . 11)
( - 2 .9 2 )
( - 3 . 19 )
-2.55
-2.44
-2.43
-2.57
-2.51
-2.65
-2.31
-2.29
-2.46
-2.46
-2.497
( - 3 .4 6 )
( - 3 .2 5)
( - 3 . 16 )
( - 3 .3 6 )
( - 3 .2 3 )
( - 3 .4 0 )
( - 2 .4 3 )
( - 2 .3 6 )
( - 2 .56 )
( - 2 .4 8 )
( - 2 .56 )
-2.41
-2.40
-2.42
-2.44
-2.28
-2.28
-2.99
-3.07
-3.06
-3.01
-2.92
( - 3 . 14 )
( - 3 .0 3 )
( - 3 .0 3 )
( - 3 .0 6 )
( - 2 .8 3 )
( - 2 .8 0 )
( - 2 .9 4 )
( - 3 .0 2 )
( - 3 .0 0 )
( - 2 .9 4 )
( - 2 .8 5)
-1.65
-1.57
-1.52
-1.56
-1.47
-1.53
-1.73
-1.703
-1.75
-1.586
-1.66
( - 2 .3 4 )
( - 2 . 17 )
( - 2 .0 8 )
( - 2 . 14 )
( - 1. 9 7 )
( - 2 .0 2 )
( - 1. 8 5 )
( - 1. 8 0 )
( - 1. 8 5 )
( - 1. 6 5 )
( - 1. 7 3 )
2.34
0.40
-0.42
-0.18
1.29
1.13
-1.04
-0.96
-0.51
1.76
0.67
( 1. 5 3 ) .
( - 0 .2 5)
( - 0 .2 6 )
( - 0 . 11)
( 0 .76 ) .
( 0 .6 5) .
( - 0 . 5 1)
( - 0 .4 7)
( - 0 .2 4 )
( 0 .72 ) .
( 0 .2 7) .
^^^
^^^
^^^
321
0.2658
14.96
1.69
321
0.2939
16.78
1.67
310
0.2907
14.78
1.70
321
0.2846
14.33
1.69
309
0.2708
16.01
1.70
^^^
320
0.2634
15.41
1.69
239
0.2940
11.76
1.68
228
0.2904
10.19
1.71
239
0.2817
9.35
1.71
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
Only Lo w inc o me
Out lie rs Omit t e d
2 S LS
Observations
R^2
F-Stat
F-Crit at 95%
0.492
( 0 .0 5) .
27.360
**
( 2 .2 4 ) .
elrpolicy4
-8.028
( - 0 .8 5)
227
0.2602
9.84
1.71
^^^
238
0.2741
9.56
1.71
78
Table 6
CPI Policy, ELR Countries and Years
FULL SAMPLE
5/OLS
5+/OLS
P o lic y 3
Variable
1
Coefficient
(t-value)
elrlgdp
4/OLS
5/2SLS
5+/2SLS
LOW-INCOME COUNTRY SAMPLE
8/OLS
8+/OLS
8/2SLS
8+/2SLS
7/OLS
0.015
-0.004
-0.072
-0.02
-0.053
0.023
0.082
0.051
0.047
0.011
0.131
( 0 .0 8 ) .
( - 0 .0 2 )
( - 0 .3 9 )
( - 0 . 11)
( - 0 .2 7)
( 0 . 12 ) .
( 0 .4 0 ) .
( 0 .2 4 ) .
( 0 .2 3 ) .
( 0 .0 5) .
( 0 .6 2 ) .
elrethnf
-0.347
-0.058
0.245
0.025
0.101
-0.454
0.529
0.691
0.708
0.975
0.264
( - 0 .4 7)
( - 0 .0 8 )
( 0 .3 6 ) .
( 0 .0 4 ) .
( 0 . 14 ) .
( - 0 .6 0 )
( 0 .73 ) .
( 0 .9 4 ) .
( 0 .9 7) .
( 1. 2 5 ) .
( 0 .3 3 ) .
-1.403
** -1.327
-0.683
-1.091
( - 2 . 18 )
( - 1. 2 8 )
( - 0 .8 9 )
( - 1. 7 1)
elrssa
elreasia
-0.87
-0.932
( - 1. 5 5 )
( - 1. 3 1)
2.551 ***
( 4 .8 8 ) .
elricrge
elrm21
2.593
-1.594 **
( - 2 . 19 )
***
( 5.0 9 ) .
0.183
0.213
( 1. 5 6 ) .
( 1. 7 6 ) .
2.53
*** 2.644 *** 2.548 *** 2.842 ***
( 4 .8 4 ) .
*
0.312
( 5 . 12 ) .
**
( 2 .53 ) .
0.233
( 4 .2 6 ) .
*
( 1. 9 4 ) .
0.281
( 4 . 14 ) .
**
( 2 . 16 ) .
2.447
*
*** 2.425
( 4 .4 9 ) .
0.165
0.272
( 1. 3 4 ) .
( 1. 7 3 ) .
-1.862 *** -1.529 ** -2.086 **
( - 2 .8 6 )
***
( 4 .3 4 ) .
*
0.431
( - 2 .2 9 )
2.64
*** 2.377 ***
2.57
( 4 .72 ) .
( 3 .3 3 ) .
( 3 .8 8 ) .
*** 0.253
0.418
( 1. 5 8 ) .
( 2 .55) .
( 2 .6 4 ) .
-1.19
( - 2 .54 )
**
0.271
0.012
0.004
0.000
0.004
0.002
0.016
0.014
0.012
0.015
0.011
0.025
( 0 .3 4 ) .
( - 0 .0 2 )
( 0 .3 8 ) .
( 0 . 17 ) .
( 1. 14 ) .
( 1. 0 2 ) .
( 0 .8 3 ) .
( 1. 0 4 ) .
( 0 .6 4 ) .
( 1. 5 6 ) .
-8.243
14.896
10.31500
38.483 **
( - 0 .3 5)
( 0 .6 9 ) .
( 0 .3 3 ) .
( 2 . 3 1) .
9.902
***
*
( 1. 7 2 ) .
( 1. 0 6 ) .
elrbb
*
( - 1. 7 8 )
**
( 2 .0 0 ) .
elrinfl
-1.141 **
( - 2 .4 6 )
elrsacw
0.002
( 0 .3 2 ) .
elraid
elraidpolicy3
elraid2policy3
15.927
13.554
33.489 ** -10.012
( 0 . 9 1) .
( 0 . 6 1) .
( 2 . 4 1) .
-14.816
** -2.546
-12.675
( - 2 . 0 1)
( - 0 . 2 1)
60.945
( - 1. 7 4 )
( - 0 .3 9 )
*
7.426
2.46
-16.992
( 0 .3 8 ) .
( 0 .2 0 ) .
( - 1. 6 9 )
***
1.3190
( 3 . 5 1) .
Period 3
Period 4
Period 5
Period 6
Period 7
Period 8
Constant
-3.196
-16.41
( - 0 . 18 )
( - 1. 7 3 )
*
5.724
( 0 . 19 ) .
-14.84
-2.697
( - 0 . 5 1)
( - 0 . 16 )
65.575 ***
( 2 .6 2 ) .
elrpolicy3
*
36.49
( 0 .8 6 ) .
( 2 .6 2 ) .
*** 1.164 *** 1.379
( 2 .9 2 ) .
( 3 .58 ) .
*** 0.898 **
0.937
( 2 .0 4 ) .
( 2 .2 9 ) .
**
1.541
***
( 2 .6 0 ) .
1.239
*
( 1. 7 7 ) .
1.727 *** 1.701
( 2 .77) .
1.139
( 1. 6 2 ) .
( 1. 3 6 ) .
0.19
-0.05
0.13
-0.05
0.23
0.27
0.10
0.25
0.08
0.16
0.27
( 0 .2 5) .
( - 0 .0 7)
( 0 . 17 ) .
( - 0 .0 6 )
( 0 . 3 1) .
( 0 .3 7) .
( 0 . 14 ) .
( 0 .3 5) .
( 0 . 11) .
( 0 .2 3 ) .
( 0 .3 6 ) .
-1.60
-1.80
-1.74
-1.82
-1.67
-1.53
-1.22
-1.08
-1.28
-1.14
-1.11
( - 2 .0 2 )
( - 2 .2 9 )
( - 2 . 10 )
( - 2 .3 0 )
( - 2 . 0 1)
( - 1. 8 5 )
( - 1. 3 2 )
( - 1. 11)
( - 1. 3 6 )
( - 1. 19 )
( - 1. 19 )
-2.96
-2.87
-2.86
-2.95
-2.87
-2.95
-2.06
-2.11
-2.18
-2.06
-2.28
( - 4 .4 4 )
( - 4 .3 6 )
( - 4 . 19 )
( - 4 .4 0 )
( - 4 .2 7)
( - 4 .4 6 )
( - 2 .58 )
( - 2 .58 )
( - 2 .6 9 )
( - 2 .53 )
( - 2 .8 3 )
-2.22
-2.03
-2.06
-2.11
-2.10
-2.24
-1.46
-1.59
-1.60
-1.57
-1.75
( - 3 .3 3 )
( - 3 .0 8 )
( - 3 .0 2 )
( - 3 . 12 )
( - 3 . 13 )
( - 3 .3 8 )
( - 1. 9 2 )
( - 1. 9 9 )
( - 2 .0 3 )
( - 1. 9 9 )
( - 2 .2 3 )
-1.97
-1.89
-1.85
-1.99
-1.77
-1.93
-2.17
-2.32
-2.35
-2.34
-2.34
( - 2 .75)
( - 2 .77)
( - 2 .52 )
( - 2 .8 5)
( - 2 .4 0 )
( - 2 .79 )
( - 2 .6 3 )
( - 2 .57)
( - 2 .78 )
( - 2 .58 )
( - 2 .8 0 )
-1.08
-0.96
-0.96
-1.04
-0.94
-1.10
-0.82
-0.87
-0.98
-0.90
-1.04
( - 1. 6 8 )
( - 1. 5 6 )
( - 1. 4 9 )
( - 1. 6 3 )
( - 1. 5 0 )
( - 1. 7 3 )
( - 1. 0 8 )
( - 1. 10 )
( - 1. 2 4 )
( - 1. 18 )
( - 1. 3 3 )
2.23
-0.36
-0.19
-0.69
0.34
0.15
-2.31
-2.28
-2.77
-2.96
-2.21
( 1. 4 8 ) .
( - 0 . 2 1)
( - 0 . 11)
( - 0 . 4 1)
( 0 .2 0 ) .
( 0 .0 9 ) .
( - 1. 19 )
( - 1. 0 9 )
( - 1. 4 4 )
( - 1. 2 1)
( - 1. 0 9 )
Only Lo w inc o me
Out lie rs Omit t e d
2 S LS
Observations 312
R^2
0.2565
F-Stat
14.55
F-Crit at 95% 1.7
^^^
^^^
312
0.2922
18.12
1.7
294
0.2834
14.30
1.73
^^^
^^^
312
0.279
13.79
1.73
293
0.2782
14.85
1.73
^^^
^^^
^^^
^^^
^^^
^^^
^^^
311
0.2411
13.7
1.73
226
0.3360
12.37
1.72
209
0.3287
9.94
1.75
226
0.3167
8.99
1.75
208
0.3226
9.20
1.75
^^^
^^^
225
0.2976
9.69
1.75
79
Table 7: GDP Deflator Policy, ELR Countries, All Years
Policy6
Variable
elrlgdp
-0.581
**
( - 3 .8 2 )
elrethnf
elrssa
***
-0.393
***
( - 2 .4 7)
-0.731
***
( - 3 .76 )
-0.556
***
( - 2 .8 8 )
-0.613
**
( - 2 .4 5)
-0.569
**
( - 2 . 2 1)
-0.532
**
( - 2 . 11)
-0.630
**
( - 2 .3 4 )
-0.495
-0.479
-0.155
-0.323
-0.712
-0.799
-0.624
-0.260
-0.484
-0.709
-0.681
( - 0 .2 9 )
( - 0 .59 )
( - 1. 15 )
( - 1. 3 1)
( - 0 .9 4 )
( - 0 .4 0 )
( - 0 .74 )
( - 0 .9 6 )
( - 0 .9 7)
-1.539
***
1.441
0.251
-0.001
9.318
-1.617
***
( - 3 .6 2 )
***
1.502
***
( 3 .6 3 ) .
***
0.278
-0.001
***
1.443
***
0.344
***
***
-0.001
***
1.615
***
( 3 .8 6 ) .
***
( 3 . 8 1) .
( - 9 .50 )
-1.933
( - 4 .53 )
( 3 .4 8 ) .
( 3 .0 2 ) .
***
-1.856
( - 4 .3 8 )
0.289
***
-0.001
-1.248
( - 2 .3 4 )
1.279
***
( 3 .0 8 ) .
***
( 3 . 12 ) .
( - 8 .4 3 )
-0.772
( - 1. 2 4 )
0.281
***
***
-0.001
0.220
***
***
-0.001
( - 7.6 4 )
***
1.845
**
0.257
***
-0.001
***
1.854
**
0.333
***
***
-0.001
***
2.011
***
( 4 .2 4 ) .
***
( 2 .6 2 ) .
( - 7.70 )
-1.821
( - 4 .0 3 )
( 3 .9 6 ) .
( 1. 9 8 ) .
***
-1.903
( - 4 .2 0 )
( 3 .9 2 ) .
( 2 .3 0 ) .
( - 7.78 )
-1.512
( - 3 . 19 )
( 3 .6 9 ) .
( 2 .9 7) .
( - 9 . 12 )
1.542
**
0.259
***
-0.001
-1.533
( - 2 . 9 1)
***
( 3 .2 6 ) .
**
( 1. 9 9 ) .
( - 7 . 10 )
-1.040
( - 1. 6 3 )
1.567
0.310
***
-0.001
1.972
***
***
( 4 .0 4 ) .
**
( 2 .4 0 ) .
( - 7.4 6 )
*
( - 1. 8 9 )
( - 0 .8 8 )
( - 10 . 14 )
elrbb
-0.477
( - 3 .0 2 )
LOW-INCOME COUNTRY SAMPLE
8/OLS
8+/OLS
8/2SLS
8+/2SLS
7/OLS
-0.568
( 2 .9 5) .
elrm21
***
5+/2SLS
( - 1. 0 4 )
( 3 . 4 1) .
elricrge
-0.512
( - 3 .2 4 )
( - 4 .2 3 )
elreasia
FULL SAMPLE
5/OLS
5+/OLS
5/2SLS
1
4/OLS
Coefficient
(t-value)
0.248
**
( 1. 9 7 ) .
***
( - 6 .58 )
-0.001
***
( - 6 . 14 )
***
( 3 .2 8 ) .
klinfl
-0.116
***
( - 4 .56 )
elrsacw
0.006
( 1. 15 ) .
elraid
elraidpolicy6
elraid2policy6
18.559
-0.332
21.128
-45.51
-39.225
41.458
25.088
45.816
-14.059
-9.822
( 0 .70 ) .
( - 0 . 0 1)
( 0 .79 ) .
( - 0 .9 9 )
( - 1. 0 2 )
( 1. 2 6 ) .
( 0 .74 ) .
( 1. 4 0 ) .
( - 0 . 2 1)
( - 0 . 18 )
-3.750
1.037
-1.507
3.23
5.384
-7.366
-2.482
-5.447
-0.312
2.438
( - 0 .8 7)
( 0 .2 3 ) .
( - 0 .3 5)
( 0 .4 0 ) .
( 0 .8 4 ) .
( - 1. 3 9 )
( - 0 .4 6 )
( - 1. 0 3 )
( - 0 .0 3 )
( 0 .2 7) .
12.462
11.910
***
( 2 .8 3 ) .
elrpolicy6
1.068
( 5 . 15 ) .
Period 1
Period 2
Period 3
Period 4
Period 5
Period 6
Period 7
Period 8
Period 9
Period 10
Period 11
Constant
***
0.998
***
( 5.0 6 ) .
1.039
( 5.0 5) .
***
0.878
***
( 3 . 6 1) .
0.830
***
( 3 .6 9 ) .
1.261
***
( 3 .9 0 ) .
1.169
***
( 3 .8 5) .
1.244
***
( 3 .9 0 ) .
0.986
**
( 1. 9 8 ) .
-1.39
-1.41
-1.42
-1.37
-1.384
-1.40
-1.40
( - 2 .2 6 )
( - 2 .2 8 )
( - 2 .3 3 )
( - 2 . 10 )
( - 2 . 2 1)
( - 1. 7 0 )
( - 1. 6 2 )
0.904
1.34
1.40
1.42
1.43
( 1. 6 8 ) .
( 1. 7 3 ) .
( 1. 7 2 ) .
0.75
-0.67
-0.84
-0.65
-0.80
-0.366
0.67
-0.80
0.72
-0.75
0.85
( 1. 5 0 ) .
( - 1. 0 3 )
( - 1. 3 3 )
( - 0 .9 9 )
( - 1. 2 2 )
( - 0 .53 )
( 1. 13 ) .
( - 0 .9 5)
( 1. 2 3 ) .
( - 0 .8 4 )
( 1. 3 5 ) .
-0.48
-1.92
-1.80
-1.92
-1.63
-1.565
-0.47
-1.66
-0.45
-1.46
-0.29
( - 0 . 8 1)
( - 2 .4 8 )
( - 2 .2 8 )
( - 2 .4 9 )
( - 2 .0 0 )
( - 1. 9 5 )
( - 0 .6 2 )
( - 1. 6 2 )
( - 0 . 6 1)
( - 1. 4 0 )
( - 0 .3 7)
-2.23
-3.37
-3.24
-3.44
-3.08
-3.188
-1.514
-2.70
-1.59
-2.53
-1.49
( - 4 .53 )
( - 4 .8 6 )
( - 4 .59 )
( - 4 .9 7)
( - 4 .2 7)
( - 4 .6 4 )
( - 2 .3 5)
( - 2 .8 7)
( - 2 .52 )
( - 2 .6 7)
( - 2 .3 6 )
-0.83
-2.09
-2.10
-2.15
-1.87
-2.02
-0.58
-2.04
-0.66
-1.77
-0.60
( - 1. 6 7 )
( - 3 .0 7)
( - 3 .0 2 )
( - 3 . 16 )
( - 2 .6 8 )
( - 2 .9 8 )
( - 0 .9 5)
( - 2 . 2 1)
( - 1. 0 8 )
( - 1. 9 6 )
( - 0 .9 9 )
-0.58
-2.04
-1.98
-2.11
-1.60
-1.817
-1.46
-2.88
-1.56
-2.45
-1.34
( - 1. 13 )
( - 2 .8 5)
( - 2 .70 )
( - 2 .9 7)
( - 2 . 15 )
( - 2 .53 )
( - 2 . 17 )
( - 2 .9 4 )
( - 2 .3 7)
( - 2 .4 5)
( - 1. 9 2 )
0.41
-1.15
-1.07
-1.18
-0.93
-1.065
0.07
-1.26
0.02
-1.06
0.10
( 0 .9 3 ) .
( - 1. 8 4 )
( - 1. 7 1)
( - 1. 8 8 )
( - 1. 4 4 )
( - 1. 7 0 )
( 0 . 12 ) .
( - 1. 4 7 )
( 0 .0 4 ) .
( - 1. 2 3 )
( 0 . 18 ) .
-0.93
-2.39
-2.36
-2.36
-2.36
-2.353
-0.71
-2.09
-0.68
-2.04
-0.66
( - 2 . 11)
( - 3 .8 6 )
( - 3 .79 )
( - 3 . 8 1)
( - 3 .6 4 )
( - 3 .72 )
( - 1. 2 7 )
( - 2 .4 8 )
( - 1. 2 6 )
( - 2 .3 7)
( - 1. 2 0 )
0.89
-0.78
-0.81
-0.73
-0.79
-0.623
0.48
-1.00
0.53
-0.92
0.61
- 2 .1
( - 1. 2 8 )
( - 1. 3 4 )
( - 1. 2 0 )
( - 1. 2 6 )
( - 1. 0 0 )
( 0 . 9 1) .
( - 1. 2 3 )
( 1. 0 3 ) .
( - 1. 11)
( 1. 17 ) .
1.03
-0.65
-0.67
-0.65
-0.63
-0.613
0.86
-0.59
0.84
-0.48
0.88
( 2 .4 4 ) .
( - 1. 0 6 )
( - 1. 0 8 )
( - 1. 0 6 )
( - 0 .9 8 )
( - 0 .9 9 )
( 1. 5 6 ) .
( - 0 . 7 1)
( 1. 5 8 ) .
( - 0 .57)
( 1. 6 4 ) .
6.70
-0.59
-0.90
-1.51
2.43
1.57
-2.49
-1.42
-3.24
0.80
-1.13
( 5.3 8 ) .
( - 0 .3 3 )
( - 0 .52 )
( - 0 .8 6 )
( 1. 12 ) .
( 0 .75) .
( - 0 .9 9 )
( - 0 .57)
( - 1. 3 1)
( 0 .2 4 ) .
( - 0 .3 7)
^^^
^^^
1.60
577
0.2996
53.18
1.59
553
0.2955
43.42
1.61
^^^
^^^
553
0.2516
64.81
1.61
553
0.2516
64.81
1.61
^^^
^^^
^^^
^^^
^^^
^^^
^^^
577
0.2525
58.54
1.61
408
0.3091
42.81
1.61
386
0.3059
58.20
1.61
408
0.2993
41.72
1.61
386
0.2707
51.93
1.61
**
( 2 .0 9 ) .
( 2 .2 3 ) .
Only Lo w inc o me
Out lie rs Omit t e d
2 S LS
Observations
R^2
F-Stat
F-Crit at 95%
**
( 2 . 5 1) .
^^^
^^^
408
0.2897
38.05
1.61
80
Table 8
CPI Policy, ELR Countries, All Years
FULL SAMPLE
5+/OLS
5/OLS
Policy5
4/OLS
1
Variable
elrlgdp
elrethnf
elrssa
Coefficient
(t-value)
** -0.293
( - 2 .2 9 )
( - 2 . 18 )
( - 1. 5 1)
( - 2 .6 6 )
( - 1. 6 3 )
( - 2 .8 8 )
( - 2 .77)
( - 2 .6 5)
( - 2 .76 )
-0.339
-0.326
-0.035
-0.096
-0.574
-0.511
-0.656
-0.400
-0.468
-0.808
-0.570
( - 0 .6 6 )
( - 0 .6 2 )
( - 0 .0 7)
( - 0 . 18 )
( - 0 .9 4 )
( - 0 .8 6 )
( - 1. 0 7 )
( - 0 .6 5)
( - 0 .77)
( - 1. 17 )
( - 0 .9 0 )
-0.396 ** -0.436 **
1.686 *** 1.710
*** 1.660
( 4 .2 2 ) .
( 2 .6 4 ) .
*** 1.810
( 4 .4 2 ) .
( 4 .0 7) .
0.215 *** 0.228 *** 0.301
***
( - 2 .3 9 )
1.512
*** 1.800 ***
1.88
( 3 .6 5) .
( 4 .2 8 ) .
( 4 . 13 ) .
( 2 .9 0 ) .
( 3 .54 ) .
**
0.217
*** 1.931
*** 2.039 *** 1.596
**
0.312 *** 0.224
**
0.277
( 4 .2 3 ) .
**
0.218
**
( 1. 9 8 ) .
( 2 .4 4 ) .
( 2 .0 2 ) .
( 2 .8 3 ) .
*** 2.022 ***
( 3 .3 6 ) .
( 4 .4 4 ) .
( 4 .2 7) .
***
( - 3 .2 4 )
( - 1. 3 1)
( - 4 . 15 )
( - 3 .9 2 )
( 1. 9 6 ) .
( 2 . 18 ) .
( 2 . 7 1) .
-1.61
-1.361 *** -1.691 *** -1.752 *** -0.799
( - 3 .0 5)
( - 0 .8 7)
*** 0.251 *** 0.239 *** 0.190
( - 2 .2 8 )
-0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 *** -0.001 ***
( - 7.9 8 )
elrbb
( - 3 .6 7)
( - 3 .0 5)
( - 2 .6 6 )
( 2 .6 5) .
elrm21
-1.226 **
-1.426 *** -1.310 *** -1.466 *** -1.726 *** -0.511
( 3 .9 7) .
elricrge
-0.714 *** -0.708 *** -0.664 *** -0.747 *** -0.617
-0.661 *** -0.400
-0.421
( - 2 .0 6 )
( - 3 .59 )
elreasia
LOW-INCOME COUNTRY SAMPLE
8+/2SLS
8/2SLS
8+/OLS
8/OLS
7/OLS
5+/2SLS
5/2SLS
6.654
( - 7.3 9 )
( - 7.3 4 )
-4.083
-28.073
23.376
-80.720 ** -38.350
( - 0 .2 0 )
( - 1. 2 1)
- 1. 0 7
( - 2 .4 5)
( - 6 .0 8 )
( - 6 .9 0 )
( - 7.76 )
( - 7 . 7 1)
( - 8 . 19 )
( - 7.2 3 )
( - 8 .3 0 )
( - 7.9 5)
*
( 1. 7 5 ) .
elrinfl
-1.698 ***
( - 4 .52 )
elrsacw
-0.01
( - 0 .3 0 )
elraid
elraidpolicy5
( - 1. 3 9 )
-2.255
7.294
-3.567
12.639
8.823
-11.03
-0.032
( 1. 16 ) .
( - 0 .6 6 )
( 1. 4 5 ) .
( 1. 2 6 ) .
( - 1. 9 0 )
( - 0 .0 0 )
( 5.0 4 ) .
( 4 .9 8 ) .
( 4 . 6 1) .
Period 4
Period 5
Period 6
Period 7
Period 8
Period 9
Period 10
Period 11
Constant
0.91
0.96
0.98
1.02
0.874
0.97
( 1. 6 8 ) .
( 1. 7 5 ) .
( 1. 7 9 ) .
( 1. 8 8 ) .
( 1. 5 4 ) .
( 1. 7 6 ) .
1.303
-7.492
( 0 .0 8 ) .
( - 0 . 6 1)
*** 1.637 ***
1.105
( 3 .72 ) .
( 1. 7 9 ) .
*
1.397
-0.58
-0.61
-0.64
-0.49
-0.63
( - 0 .8 0 )
( - 0 .8 5)
( - 0 .8 9 )
( - 0 .6 6 )
( - 0 .8 9 )
1.45
1.04
0.92
1.11
0.91
1.33
0.01
-0.05
0.00
0.16
0.11
( 1. 7 2 ) .
( 1. 6 0 ) .
( 1. 8 7 ) .
( 1. 5 1) .
( 2 . 18 ) .
( 0 . 0 1) .
( - 0 .0 7)
( - 0 .0 0 )
( 0 .2 2 ) .
( 0 . 14 ) .
-0.45
-0.62
-0.56
-0.58
-0.51
-0.33
-1.10
-0.98
-1.15
-0.71
-1.03
( - 0 .70 )
( - 0 .9 7)
( - 0 .8 5)
( - 0 .9 2 )
( - 0 .72 )
( - 0 .4 8 )
( - 1. 3 0 )
( - 1. 0 9 )
( - 1. 3 4 )
( - 0 .79 )
( - 1. 17 )
-1.92
-1.89
-1.87
-1.93
-1.76
-1.82
-1.96
-1.98
-2.10
-1.66
-2.05
( - 3 .8 6 )
( - 3 .74 )
( - 3 .6 5)
( - 3 .9 0 )
( - 3 .2 4 )
( - 3 .59 )
( - 2 .70 )
( - 2 .6 3 )
( - 2 .8 4 )
( - 2 .2 4 )
( - 2 .79 )
-0.73
-0.68
-0.59
-0.70
-0.42
-0.67
-1.18
-1.13
-1.32
-0.71
-1.28
( - 1. 4 5 )
( - 1. 3 2 )
( - 1. 16 )
( - 1. 3 8 )
( - 0 .79 )
( - 1. 3 1)
( - 1. 6 3 )
( - 1. 5 2 )
( - 1. 7 8 )
( - 1. 0 0 ) .
( - 1. 7 5 )
-0.54
-0.45
-0.28
-0.54
-0.04
-0.34
-1.92
-1.88
-2.15
-1.42
-2.03
( - 1. 0 3 )
( - 0 .8 5)
( - 0 .4 9 )
( - 1. 0 1)
( - 0 .0 6 )
( - 0 .6 2 )
( - 2 . 5 1)
( - 2 .2 7)
( - 2 .76 )
( - 1. 6 6 )
( - 2 . 5 1)
0.42
0.45
0.48
0.45
0.56
0.49
-0.30
-0.35
-0.41
-0.03
-0.38
( 0 .9 6 ) .
( 1. 0 1) .
- 1. 11
( 1. 0 4 ) .
( 1. 19 ) .
( 1. 13 ) .
( - 0 .4 3 )
( - 0 .50 )
( - 0 .58 )
( - 0 .0 5)
( - 0 .55)
-0.93
-0.90
-0.87
-0.83
-0.94
-0.87
-1.22
-1.24
-1.25
-1.11
-1.25
( - 2 . 13 )
( - 2 .0 0 )
( - 1. 9 9 )
( - 1. 9 1)
( - 1. 9 5 )
( - 1. 9 7 )
( - 1. 7 8 )
( - 1. 7 8 )
( - 1. 8 0 )
( - 1. 6 6 )
( - 1. 8 1)
0.71
0.65
0.73
0.76
0.73
0.80
-0.13
-0.10
-0.12
0.11
-0.08
( 1. 7 2 ) .
( 1. 5 2 ) .
( 1. 7 5 ) .
( 1. 8 5 ) .
( 1. 6 1) .
( 1. 9 0 ) .
( - 0 .2 0 )
( - 0 . 16 )
( - 0 . 18 )
( 0 . 18 ) .
( - 0 . 12 )
0.86
0.92
0.90
0.97
0.87
0.95
0.42
0.33
0.35
0.52
0.37
( 1. 9 9 ) .
( 2 . 12 ) .
( 2 .0 7) .
( 2 . 3 1) .
( 1. 8 5 ) .
( 2 .2 2 ) .
( 0 .6 2 ) .
( 0 .4 7) .
( 0 .52 ) .
( 0 .77) .
( 0 .53 ) .
4.51
0.13
0.05
-1.55
3.10
1.01
0.93
1.42
-0.36
2.69
0.40
( 3 .2 2 ) .
( 0 .0 8 ) .
- 0 .0 3
( - 0 .9 4 )
( 1. 5 0 ) .
( 0 .4 9 ) .
( 0 . 4 1) .
( 0 .6 2 ) .
( - 0 . 15 )
( 0 .9 3 ) .
( 0 . 16 ) .
568
0.2721
48.35
1.61
559
0.2971
58.01
1.59
532
0.295
64.27
1.61
559
0.2824
58.74
1.61
532
0.2566
69.69
1.61
^^^
559
0.2494
34.55
1.61
387
0.3468
34.40
1.6
362
0.3417
48.72
1.62
387
0.3307
33.90
1.62
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
362
0.2985
41.63
1.62
**
( 2 .4 6 ) .
( 1. 9 0 ) .
Only Lo w inc o me
Out lie rs Omit t e d
2 S LS
Observations
R^2
F-Stat
F-Crit at 95%
-13.94 **
( - 2 .0 2 )
( 3 . 11) .
( 3 .72 ) .
( 4 . 18 ) .
( 4 . 14 ) .
Period 1
Period 3
( 0 .70 ) .
( - 0 . 5 1)
26.09 ***
1.0320 *** 0.884 *** 1.090 *** 0.748 *** 0.778 *** 1.485 *** 1.192
elrpolicy5
34.421
63.208 ** -31.280
( 2 .3 2 ) .
( 3 .2 0 ) .
( 3 .58 ) .
Period 2
3.117
( 0 . 11) .
( - 0 .4 6 )
28.740 ***
elraid2policy
30.925
( 1. 3 1) .
^^^
387
0.3259
30.38
1.62
81
APPENDIX C
SIGN TABLE
GDP Deflator Policy
Aid, Aid*Policy, Aid^2*Policy OLS
Panel OLS 2SLS
Panel 2SLS GMM
Aid
(-) n/s (+)**
(-) n/s (-) n/s
(-) n/s
Panel 2SLS GMM
(-) *
(-)**
Aid*Policy
(-) n/s
(+) n/s (-)*
(+) n/s
(-) n/s
(+) n/s
(+) n/s (-) n/s
(+) n/s
Aid^2*Policy
(+)*** (-)**
(+) n/s (+)*
(+) n/s
(+)***
(-) n/s
(+) n/s (+) n/s
(+) n/s
Policy
(+)*** (+)***
(+)*** (+)***
(+)***
(+)***
(+)***
(+)*** (+)***
(+)***
Aid, Aid*Policy
Aid
Aid*Policy
(-) n/s
CPI Policy
OLS
Panel OLS 2SLS
(+) n/s (+) n/s
(-)**
GDP Deflator Policy
OLS
Panel OLS 2SLS
Panel 2SLS GMM
(-) n/s (+) n/s
(-) n/s (+) n/s
(-) n/s
CPI Policy
OLS
Panel OLS 2SLS
(-) n/s (-) n/s
(-)**
(+) n/s (-) n/s
(+) n/s (-) n/s
(+) n/s
(+) n/s (+) n/s
(+) n/s (-) n/s
(+) n/s
(+)*** (+)***
(+)*** (+)***
(+)***
(+)***
(+)*** (+)***
(+)***
Panel 2SLS GMM
(-) n/s
(-)***
Aid^2*Policy
Policy
(+)***
GDP Deflator Policy
OLS
Panel OLS 2SLS
Panel 2SLS GMM
(-) n/s (+)***
(-) n/s (-) n/s
(-)*
CPI Policy
OLS
Panel OLS 2SLS
(-) n/s
(+) n/s
(-)**
Aid^2*Policy
(+) n/s (-)**
(+) n/s (+) n/s
(+) n/s
(+) n/s (+)***
(+) n/s (+) n/s
(+) n/s
Policy
(+)*** (+)***
(+)*** (+)***
(+)***
(+)***
(+)*** (+)***
(+)***
Aid
GDP Deflator Policy
OLS
Panel OLS 2SLS
(+) n/s (+)**
(-)**
Aid, Aid^2*Policy
Aid
Panel 2SLS GMM
(-) *
(-)**
Aid*Policy
Aid
Panel 2SLS GMM
(-) n/s
(-)***
(+)***
CPI Policy
OLS
Panel OLS 2SLS
(+)***
(+) n/s
(-)**
Panel 2SLS GMM
(-) n/s
(-)***
Aid*Policy
Aid^2*Policy
Policy
Aid*Policy
(+)*** (+)***
(+)*** (+)***
(+)***
(+)***
(+)***
(+)*** (+)***
(+)***
GDP Deflator Policy
OLS
Panel OLS 2SLS
Panel 2SLS GMM
CPI Policy
OLS
Panel OLS 2SLS
Panel 2SLS GMM
(+) n/s (+)**
(-)*
(-) n/s
(-)***
(+) n/s (+) n/s
(-) *
(-) n/s
(+)*** (+)***
(+)*** (+)***
(+)***
(+)***
(+)*** (+)***
Aid
Aid*Policy
(-)***
Aid^2*Policy
Policy
GDP Deflator Policy
OLS
Panel OLS 2SLS
Panel 2SLS GMM
Aid^2*Policy
(+) n/s (+) n/s
(-)*
(+) n/s
Policy
(+)*** (+)***
(+)*** (+)***
Aid^2*Policy
Aid
(+)***
(+)***
CPI Policy
OLS
Panel OLS 2SLS
Panel 2SLS GMM
(-)***
(+) n/s (+) n/s
(-) *
(+) n/s
(+)***
(+)***
(+)*** (+)***
Aid*Policy
(+)***
(-)***
(+)***
82
APPENDIX C
FINAL RESULT TABLE
FIXED EFFECTS PANEL ORDINARY LEAST SQUARES
Variable
1
Coefficient
(t-value)
2
3
4
5
6
FIXED EFFECTS PANEL 2SLS
7
Coefficient
(Z-value)
8
9
10
11
12
Log Initial GDP
SSA
EAsia
ICRGE
M2
-0.001
***
( - 4 .9 4 )
Aid
63.59
**
( 2 .2 8 ) .
Aid*Policy
Aid^2*Policy
-0.001
***
( - 4 .9 2 )
***
39.21
62.32
25.28
( 3 .0 5) .
( 2 . 13 ) .
-0.001
-2.37
3.93
( 1. 9 9 ) .
**
1.09
***
1.07
***
( 4 . 4 1) .
-0.001
***
( - 5.8 7)
-0.001
64.30
-18.88
( - 0 .73 )
( - 0 . 4 1)
379.01
( 1. 6 9 ) .
1.01
***
0.89
*
( - 1. 7 6 ) .
26.07
( 5.3 4 ) .
***
( 4 . 19 ) .
0.93
***
( 4 .9 4 ) .
-0.001
( - 2 .3 4 )
2.22
( 1. 4 5 ) .
***
-0.001
( - 1. 2 2 )
( 0 .0 4 ) .
-56.57
1.08
***
( - 0 .73 )
-32.99
**
-0.002
( - 2 . 4 1)
-70.330
( 2 .0 7) .
( 5.76 ) .
*
( - 1. 6 6 )
**
( - 0 .53 )
-56.28
***
( - 5.0 5)
-0.25
1.48
***
-0.002
-9.34
-9.16
( - 1. 0 0 )
1.29
***
( 4 .2 3 ) .
***
( - 2 .6 0 )
( - 0 . 9 1)
*
( 3 .8 4 ) .
***
-0.001
( - 0 .9 2 )
136.61
66.69
( .9 3 ) .
( 0 . 8 1) .
0.83
***
( 3 . 18 ) .
1.01
***
( 4 .8 5) .
1.29
***
( 4 .3 0 ) .
0.92
-6.02
-5.37
-5.98
-4.97
-4.25
-4.30
-3.56
-4.32
-2.22
-3.37
-4.27
-3.62
( - 3 .9 2 )
( - 5 . 19 )
( - 4 .59 )
( - 3 .70 )
( - 4 .3 0 )
( - 1. 6 5 )
( - 2 .2 4 )
( - 1. 2 6 )
( - 2 .2 2 )
( - 3 .0 5)
( - 3 .0 0 )
0.2442
0.0205
0.0813
-0.3508
0.4171
2.13
0.0000
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
^^^
0.2391
0.0012
0.1208
-0.2124
0.3672
2.08
0.0000
0.2442
0.0207
0.0809
-0.3521
0.4181
2.13
0.0000
0.2388
0.0016
0.1189
-0.2184
0.3700
2.08
0.0000
0.2370
0.0007
0.1225
-0.2036
0.3652
2.06
0.0000
0.2311
0.0225
0.1662
-0.0641
0.3190
2.03
0.0000
^^^
^^^
^^^
.. .. ..
0.1750
0.1040
-0.6466
0.4167
1.57
0.0057
^^^
^^^
^^^
0.0811
0.1482
0.1389
-0.3293
0.3167
1.86
0.0002
^^^
^^^
^^^
0.1820
0.1266
0.1776
0.0006
0.2741
2.05
0.0000
^^^
^^^
^^^
0.1889
0.1834
0.2016
0.0295
0.2626
2.11
0.0000
^^^
^^^
^^^
0.0765
0.1475
0.1370
-0.3404
0.3201
1.87
0.0002
***
( 3 .8 8 ) .
( - 4 .2 4 )
Only Lo w inc o me
Out lie rs Omit t e d
2 S LS
Time D ummie s
R^2: within
R^2: between
R^2: overall
Corr (u_i,Xb)
rho (variance due to u_i)
F test that all u_i=0
P-value
-0.001
( - 4 .9 8 )
( - 0 .0 6 )
( 4 .4 5) .
Constant
***
( 1. 4 6 ) .
( - 2 .0 0 )
Policy
-0.001
( - 5.0 0 )
^^^
^^^
^^^
0.2102
0.0026
0.9660
-0.2765
0.3871
2.19
0.0000
83
BIBLIOGRAPHY
Adelman, Irma, and Chenery, Hollis B. “The Foreign Aid and Economic Development:
The Case of Greece.” The Review of Economics and Statistics 48.1 (1966): 1-19
Addison, Tony, George Mavrotas, and Mark McGillivray. “Development Assistance and
Development Finance: Evidence and Global Policy Agendas.” Journal of
International Development 17.6 (2005): 819-836.
Aghion, Philippe, and Peter Howitt. “A Model of Growth through Creative Destruction.”
Econometrica 60.3 (1992): 23-51.
Alesina, Alberto and David Dollar. 2000. “Who Gives Foreign Aid to Whom and Why?”
Journal of Economic Growth 5.1 (2000): 33 – 63.
Alesina, Alberto and Beatrice Weder. “Do Corrupt Governments Receive Less Foreign
Aid?” The American Economic Review 92.4 (2002): 1126 – 1137.
Ali, Abdiweli and Hodan Isse. “An Empirical Analysis of the Effects of Aid on Growth.”
Internationl Advances in Economic Research 11.1 (2005): 1-11.
Arrow, Kenneth J. 1962. “The Economic Implications of Learning by Doing.” Review of
Economic Studies 29.3 (1962): 155-173.
Barro, Robert and Sala-I Martin. 1995. Economic Growth. McGraw-Hill: Boston, MA
(1995).
Benhabib, Jess, and Mark M. Spiegel. “The Role of Human Capital in Economic
Development: Evidence from Aggregate Cross-Country Data.” Journal of Monetary
Economics 34.2 (1994): 143-173.
84
Bils, Mark, and Peter Klenow. “Does Schooling Cause Growth?” American Economic
Review 90 (2000): 1160-83.
Birdsall, Nancy and Juan Luis Londono. Asset Inequality Matters: An Assessment of the
World Bank’s Approach to Poverty Reduction. The American Economic Review 87: 2
(1997), 32 – 37.
Boone, Peter. “Politics and the Effectiveness of foreign aid.” European Economic Journal
40.2 (1996): 289-329.
Bornschier, Volker, Christopher Chase-Dunn and Richard Rubinson. “Cross-national
Evidence of the Effects of Foreign Investment and Aid on Economic Growth and
Inequality: A Survey of Findings and a Reanalysis.” American Journal of Sociology
84.3 (1978): 651-683.
Bureau of Labor Statistics. “Consumer Price Index Frequently Asked Questions”
October 2011. Accessed July 2013. < http://www.bls.gov/cpi/cpifaq.htm >
Bureau of Labor Statistics. “CPI Inflation Calculator” Accessed March-July 2013.
<http://www.bls.gov/data/inflation_calculator.htm >
Burnell, Peter. “The Changing Politics of Foreign Aid-Where to Next?” Politics 17.2
(1997): 117-125.
Burnside, Craig and David Dollar. “Aid, Policies and Growth.” World Bank Policy
Research Working Paper 1777 (1997).
Burnside, Craig and David Dollar. “Aid, Policies and Growth.” American Economic
Review 90.4 (2000): 847-868.
85
Cabelle, Jordi, and Manuel S. Santos. “On Endogenous Growth with Physical and Human
Capital.” Journal of Political Economy 101. 6 (1993): 1042-1067.
Cass, David. “Optimum Growth in an Aggregative Model of Capital Accumulation.”
Review of Economic Studies 32.3 (1965): 233-240.
Chandra, Siddharth. “Aid and Development: A Conceptual Perspective from
Development Economics.” In Foreign Aid and Foreign Policy: Lessons for the Next
Half-Century. Ed Louis A Picard, Robert Groelsema, and Terry F Buss. M.E. Sharpe,
Inc: Armonk, NY (2008).
Chang, Charles, Eduardo Fernández-Arias, and Luis Servén. “Measuring Aid Flows: A
New Approach.” World Bank Policy Research Working Paper 2050 (1998).
Chang, Charles, Eduardo Fernández-Arias, and Luis Servén. “Dataset for: Measuring Aid
Flows: A New Approach, 1998” Accessed March 2013
<http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH/0,,co
ntentMDK:20827466~pagePK:64214825~piPK:64214943~theSitePK:469382,00.htm
l>
Chang, Ha-Joon. Bad Samaritans: The Myth of Free Trade and the Secret History of
Capitalism. Randon House: New York, NY (2008).
Chatterjee, Santanu, Paulo Giuliano and Ilker Kaya. “Where Has All the Money Gone?
Foreign Aid and the Quest for Growth.” Institute for the Study of Labor Discussion
Paper 2858, June 2007.
Chenery, Hollis B. and Alan M. Strout. “Foreign Assistance and Economic
Development.” American Economic Review 56.4 (1966): 679-733.
86
Chenery, Hollis B., Montek S. Ahluwalia, and Nicholas G Carter. “Growth and Poverty
in Developing Countries.” Journal of Development Economics 6.3 (1979): 299-341.
Cogneau, Denis and Jean-David Naudet. “Who Deserves Aid? Equity of Opportunity,
International Aid, and Poverty Reduction.” World Development 35.1 (2007): 104120.
Cohen, Jesicca, and William Easterly. 2009. “Introduction: Thinking Big versus Thinking
Small.” In What Works in Development?: Thinking Big and Thinking Small. Eds:
Cohen and Easterly. Brookings Institute: Washington, DC (2009).
Collier, Paul and David Dollar. “Can the World Cut Poverty in Half? How Policy Reform
and Effective Aid Can Meet International Development Goals.” World Development
29.11 (2001): 1787-1802.
Collier, Paul and David Dollar. 2002. “Aid Allocation and Poverty Reduction.” European
Economic Review 46.8 (2002): 1475-1500.
Collier Paul and Jan Dehn. “Aid, Shocks and Growth.” World Bank Policy Research
Working Paper 2688 (2001).
Collier, Paul and Anke Hoeffler. “Aid, Policy and Growth in Post-Conflict Societies.”
World Bank Policy Research Working Paper 2902 (2002).
Dalgaard, Carl-Johan and Henrik Hansen. “On Aid, Growth and Good Policy.” Journal of
Development Studies 37.6 (2001): 17-41.
Dalgaard, Carl-Johan, Henrik Hansen and Finn Tarp. 2004. “On the Empirics of Foreign
Aid and Growth.” The Economic Journal 114.496 (2004): F191-F216.
87
Dengnbol-Martinussen, John and Poul Engberg-Pedersen. Aid: Understanding
International Development Cooperation. Zed Books Ltd: New York, NY (2003).
Dollar, David and Victoria Levin. “Sowing and Reaping: Institutional Quality and Project
Outcomes in Developing Countries.” World Bank Policy Research Working Paper
3524 (2005).
Dollar, David and Victoria Levin. 2006. “The Increasing Selectivity of Foreign Aid,
1984-2003.” World Development 34.12 (2006): 2034-2046.
Durbarry, Ramesh, Norman Gemmell and David Greenaway. “New Evidence on the
Impact of Foreign Aid on Economic Growth.” Center for Research in Economic
Development and International Trade, University of Nottingham. CREDIT Research
Paper 98/8 (1998).
Easterly, William. The White Man’s Burden. The Penguin Press: New York, NY (2006).
Easterly, William, Ross Levine and David Roodman. “New Data, New Doubts: A
Comment on Burnside and Dollar’s ‘Aid, Policies, and Growth’ (2000).” National
Bureau of Economic Research Working Paper 9846 (2003).
Easterly, William, Ross Levine and David Roodman. “New Data, New Doubts:
Revisiting ‘Aid, Policies, and Growth’ Working Paper 26 Data and Stata 7 command
file. Accessed March 2013 <http://www.cgdev.org/publication/new-data-new-doubtsrevisiting-aid-policies-and-growth-working-paper-26>
Fayissa, Bichaka, and Mohammed El-Kaissy. “Foreign aid and the economic growth of
developing countries (LDCs): Further evidence.” Studies in Comparative
International Development 34.3 (1999): 37-50.
88
Friedman, Milton. “Foreign Economic Aid: Means and Objectives.” The Yale Review 47
(1958): 500-516.
Friedman, Milton. Capitalism and Freedom. Chicago: University of Chicago Press
(1962).
Giles, Judith. “Another Look at the Evidence on Foreign Aid Led Economic Growth.”
Applied Economic Letters 1.11 (1994): 194-199.
Gomanee, Karuna, Sourafel Girma and Oliver Morrissey. “Aid and Growth in SubSaharan Africa: Accounting for Transmission Mechanisms.” Center for Research in
Economic Development Research Paper 02/05 (2002).
Gounder, Rukmani. “Aid-growth nexus: Empirical Evidence from Fiji.” Applied
Economics 33.8 (2001): 1009-1019.
Griffin, K.B. and Enos, J.L. 1970. “Foreign Assistance: Objectives and Consequences.”
Economic Development and Cultural Change 18.3 (1970): 313-327.
Guillaumont, Patrick and Lisa Chauvet.. “Aid and Performance: A Reassessment.”
Journal of Development Studies 37.6 (2001): 66-92.
Hadi, Ali. 1992. “Identifying Multiple Outliers in Multivariate Data.” Journal of the
Royal Statistical Society. Series B (Methodology) 54.3 (1992): 761-771.
Hansen, Henrik and Finn Tarp. “Aid Effectiveness.” Journal of International
Development 12.3 (2000): 375-398.
Hansen, Henrik and Finn Tarp. “Aid and Growth Regressions.” Journal of Development
Economics 64.2 (2001):547-570.
89
Hermes, Niels and Robert Lensink. 2001. “Changing the Conditions for Development
Aid: A New Paradigm?” Journal of Development Studies 37.6 (2001): 1-16.
Heston, Alan, Robert Summers and Bettina Aten. “Penn World Table Version 7.1.”
Center for International Comparisons of Production, Income and Prices at the
University of Pennsylvania, July 2012. Accessed March 2013-June 2013.
<https://pwt.sas.upenn.edu/php_site/pwt71/pwt71_form_test.php>
Hiatt, Shon R and Warner P Woodworth. “Alleviating poverty through microfinance:
Village banking comes to Central America.” The Social Science Journal 43 (2006),
471 – 477.
Hook, Steven.. National interest and foreign aid. Lynne Rienner Publishing: Boulder, CO
(1995).
Huber, Evelyne, Dietrich Rueschemeyer, and John D. Stephens. 1993. “The Impact of
Economic Development on Democracy.” Journal of Economic Perspectives 7.3
(1993): 71-85.
Hudson, John and Paul Mosley. “Aid Policies and Growth: In Search of the Holy Grail.”
Journal of International Development 13.7 (2001): 1023-1038.
International Monetary Fund. “World Economic Outlook Database, 2013” Accessed
March-June 2013. <
http://www.imf.org/external/pubs/ft/weo/2013/01/weodata/index.aspx>
Islam, Anisul. “Foreign Aid and Economic Growth: an economic study of Bangladesh.”
Applied Economics 24.5 (1992): 541-544.
90
Islam, Muhammed. 2003. “Political Regimes and the Effects of Foreign Aid on
Economic Growth.” The Journal of Developing Areas 37.1 (2003): 35-53.
Jones, Charles. “R&D-Based Models of Economic Growth.” Journal of Political
Economy 103 (1995): 759-784.
Knack, Stephen. “Aid Dependence and the Quality of Governance: A Cross-Country
Empirical Analysis.” Southern Economic Journal (2000), 310-329.
Koopmans, Tjalling C. “On the Concept of Optimal Economic Growth.” In The
Econometric Approach to Development Planning. Amsterdam: North Holland (1965).
Krugman, Paul. “A Model of Innovation, Technology Transfer, and the World
Distribution of Income.” Journal of Political Economy 87.2 (1979): 253-266.
Lensink, Robert and Howard White. “Are there Negative Returns to Aid?” Journal of
Develpoment Studies 37.6 (2001): 42-65.
Lucus, Robert E., Jr. “On the Mechanics of Economic Development.” Journal of
Monetary Econometrics 22.1 (1988): 3-42.
Mankiw, N. Gregory, David Romer, and David N Weil. “A Contribution to the Empirics
of Economic Growth.” Quarterly Journal of Economics 107.2 (1992): 407-437.
Maren, Michael. The Road to Hell: the Ravaging Effects of Foreign Aid and International
Charity. The Free Press: New York, NY (1997).
Mbaku, John. “Foreign aid and Economic Growth in Cameroon.” Applied Economics
25.10 (1993): 1309-1314.
91
Minoui, Camelia and Sanjay Reddy. “Development Aid and Economic Growth: A
Positive Long-Run Relation.” Quarterly Review of Economics and Finance 50.1
(2010): 27-39.
Morrissey, Oliver. “Does Aid Increase Growth?” Progress in Development Studies 1.1
(2001): 37-50.
Mosley, Paul. “Aid, Savings and Growth Revisited.” Oxford Bulletin of Economics and
Statistics 42.2 (1980): 79-95.
Mulligan, Casey B., and Xavier Sala-i-Martin. “Transitional Dynamics in Two-Sector
Models of Endogenous Growth.” Quarterly Journal of Economics 108.3 (1993): 737773.
Nelson, Richard R., and Edmund S. Phelps. “Investment in Humans, Technological
Diffusion, and Economic Growth.” American Economic Review 56.2 (1966): 69-75.
Neumayer, Eric. “Do Human Rights Matter in Bilateral Allocation? A Quantitative
Analysis of 21 Donor Countries.” Social Science Quarterly 84.3 (2003): 650 – 666.
Organization for Economic Co-operation and Development. “Aid (ODA) Disbursements
to countries and regions [DAC2a] Table in constant million 2011 US$, 2012”
Accessed April 2013 <http://stats.oecd.org/index.aspx>
Organization for Economic Co-operation and Development. “DAC List of Aid Recipients
Used for 1997, 1998, and 1999 Flows.” Accessed July 2013.
<http://www.oecd.org/dac/stats/1809409.htm >
92
Organization for Economic Co-operation and Development. “DAC List of Aid
Recipients- As of 1 Juanuary 2003” Accessed July 2013. <
http://www.oecd.org/dac/stats/2488552.pdf>
Papanek, Gustav. “The Effect of Aid and Other Resource Transfers on Savings and
Growth in Less Developed Countries.” The Economic Journal 82.327 (1972): 934950.
Papanek, Gustav. “Aid, Private Foreign Investment, Savings and Growth in Less
Developed Countries.” The Journal of Political Economy 81.1 (1973): 120-130.
Pietrobelli, Carlo and Carlo Scarpa. “Inducing Efficiency in the Use of Foreign Aid: The
Case for Incentive Mechanisms.” The Journal of Development Studies 29:1 (1992),
72 – 92.
Politonomist, The. “GDP Deflator and Measuring Inflation.” January 11, 2009. Accessed
July 2013. < http://www.politonomist.com/gdp-deflator-and-measuring-inflation00491/>
Radelat, Steven, Michael Clemens and Rikhil Bhavnani.“Aid and Growth.” Finance and
Development 42.3 (2005).
Rajan, Raghuram G. and Subramanian, Arvind. “What Undermines Aid's Impact on
Growth?” International Monetary Fund, IMF Working Paper 05/126 (2005).
Rajan, Raghuram G. and Subramanian, Arvind. “Aid and Growth: What Does the CrossCountry Evidence Really Show?” Review of Economics and Statistics 90.4 (2008):
643-665.
93
Ramsey, Frank. “A Mathematical Theory of Savings.” Economic Journal 38.152 (1928);
543-559.
Romer, Paul M. “Increasing Returns and Long-Run Growth.” Journal of Political
Economy 94.5 (1986): 1002-1037.
Romer, Paul M. “Growth Based on Increasing Returns Due to Specialization.” American
Economic Review 77.2 (1987): 56-62.
Romer, Paul M. “Endogenous Technological Change.” Journal of Political Economy 98.5
(1990): pt II, S71-S102.
Roodman, David.“The Anarchy of Numbers: Aid, Development, and Cross-Country
Empirics.” Center for Global Development (2004).
Segerstrom, Paul S. “Innovation, Imitation, and Economic Growth.” Journal of Political
Economy 99.4 (1991): 807-827.
Sheshinski, Eytan. “Optimal Accumulation with Learning by Doing.” In Karl Shell, ed.,
Essays on the Theory of Optimal Economic Growth, 31-52. Cambridge, MA: MIT
Press (1967).
Shultz, T. Paul. “Returns to Women’s Education,” in Women’s Education in Developing
Countries: Barriers, Benefits, and Policies Ed by Elizabeth M King and M. Anne Hill.
Baltimore, MD: The Johns Hopkins University Press. Pp51-99 (1993).
Singh, Ram. “State Intervention, Foreign Economic Aid, Savings, and Growth in LDCs:
Some Recent Evidence.” Kyklos 38,2: 216-232 (1985).
Solow, Robert M. “A Contribution to the Theory of Economic Growth.” The Quarterly
Journal of Economics 70.1 (1956): 65-94.
94
Snyder, Donald. 1990. “Foreign Aid and Domestic Savings: A Spurious Correlation.”
Economic Development and Cultural Change39.1 (1990): 175-181.
Swan, Trevor. “Economic Growth and Capital Accumulation.” Economic Record 32.2
(1956): 334-361.
United Nations Conference on Trade and Development (UNCTAD). “UNCTAD
Statistical Database.” Accessed March-June 2013.
<http://unctadstat.unctad.org/ReportFolders/reportFolders.aspx>
Uzawa, Hirofumi. “Optimal Technical Change in an aggregative Model of Economic
Growth.” International Economic Review 6.1 (1965): 18-31.
Wallace, Tina, Lisa Bornstein and Jennifer Chapman. The Aid Chain: Coercion and
Commitment in Development NGOs. Practical Action Publishing: Rugby, UK (2007)
Shleifer, Andrei and Robert Vishny. “Corruption.” The Quarterly Journal of Economics
108.3 (1993): 599 – 617.
World Bank. “Assessing Aid: What Works, What Doesn’t, and Why.” World Bank:
Washington, DC (1998).
United Nations. “Practical Guide to Producing Consumer Price Indicies.” United Nations:
New York (2009).
United Nations Statistics Division. “National Accounts Division of Main Aggregates
Database.” Accessed June 2013. <http://unstats.un.org/unsd/snaama/selbasicFast.asp>
World Bank. “World Development Indicators Database, 2013” Accessed February-July
2013 <http://data.worldbank.org/data-catalog/world-development-indicators >
95
World Trade Organization. “Total Merchandise Trade Dataset.” Accessed June 2013.
<http://stat.wto.org/StatisticalProgram/WSDBStatProgramSeries.aspx?Language=E>
Yunus, Muhammad, and Alan Jolis. Banker to the Poor: Micro-lending and the Battle
Against Poverty. New York: PublicAffairs (2003).
Zanger, Sabine. “Good Governance and European Aid: The Impact of Political
Conditionality.” European Union Politics 1.3 (2000); 293 – 317.