By: Eric Bruce
Following
World War II, Official Development Assistance - more commonly discussed as
“International Aid” - has been a feature of the foreign policy of western
nations towards the developing world. Economic
theory suggests that financial inflows to developing nations may help to “jump
start” lagged economies and improve the welfare of their citizens. Detractors
of aid programs sometimes claim that aid money is only financing corrupt and
weak foreign governments that divert funds away from their intended purpose.
Understanding whether or not the presence of local corruption undermines the
effectiveness of international aid is important for policy makers faced with
the task of determining where to allocate scarce aid funding. This study
examines aid recipient countries exhibiting various levels of public-sector
corruption and looks for differences in the relationships between human
development and aid spending. I find evidence that corruption
may be a significant factor in lessening the impact of international aid.[1]
Study Design
Significant research has been
conducted on the topic of aid effectiveness with equally significant
disagreement on the results. The myriad long-term goals of international aid along
with the society level scale of the desired results make isolating the exact
effects of aid very difficult. Traditional measures of aid effectiveness, like
GDP per Capita, only address a small piece of the puzzle. The UN Human
Development Index (HDI) provides a partial solution to this problem by
addressing several key components of development beyond simple economic
capacity. In particular, the HDI includes measures of education, life
expectancy, and living standards (among others). This study uses HDI has a
proxy for overall human welfare and examines correlations between aid inflows
over time to changes in HDI.[2]
In order to test a
hypothesis regarding the impact of corruption on the effectiveness of aid, a
model must be developed that describes changes in HDI over time as a function
of foreign aid. Given the proposition that the HDI is a reasonable proxy
measure of human development we can utilize a standard multiple regression
model and an Ordinary Least Squares (OLS) approximation to provide insight into
the relationship between the two variables. Confounding factors are included in
the model to attempt to isolate the effect of aid from other variables that can
affect quality of life. Once a reasonable approximation of the model is
established, a corruption dummy variable (indicating a positive or negative
trend in control of government corruption) is used to assess the impact of corruption
on the examined cases.
Data were collected from multiple sources
including the World Bank Worldwide Governance Indicators, the World Bank World
Development Indicators, and the United Nations Development Program. The data
were compared for all countries for which information was available for the
time period of 2003-2012. In order to limit the challenges of longitudinal data
the dependent variable was chosen to be the total
change in HDI over the examined time period. Similarly, international aid
and corruption are measured by the sum total of values from 2003-2011. Countries
were grouped by those showing either worsening or improving corruption over the
examined time period. The distribution of aid was found to be highly skewed
with a significant number of data points at the low end and only a handful at
the upper range as can be seen in Figure 1. A log transform on aid was utilized
to approximate a normal distribution.
Results
Does aid have a positive impact on human development and, if
so, does it have a stronger or weaker impact depending on governance capacity
of recipient countries? To test this, the model described above was used to
approximate the relationship between aid and HDI accounting for confounds. The
regression results are given by equation (2) in Table 2 below. The results
demonstrate a positive and significant relationship between aid and human
development over the time period studied. Figure 2 depicts this approximately
linear relationship graphically. This shows that a 1% change in the total
amount of aid provided between 2003 and 2012 is correlated with a measurable
increase in the rate of change of
HDI.
Equation
(3) in Table 1 provides the results of the statistical analysis when the
corruption variable is included. It is important
to note that, when worsening levels of corruption are present, the coefficient
on aid (representing the effectiveness of aid) is still positive but is
decreased by an order of magnitude to the point where increases in aid
represent only a negligible effect on development. For governments that
demonstrate improving corruption control over the examined time period the
coefficient on aid is positive and doubled in magnitude from the aggregate
case. This suggests that corruption has a significant dampening effect on the
effectiveness of International Aid.
Figure 3 below depicts the effect of corruption graphically.
Discussion
The question of international aid
effectiveness is a complex one with many moving parts. The analysis conducted
above suggests a strong relationship between aid and corruption with
significant policy implications for decisions on the allocation of aid. This
study does not, however, address the potential impact that aid has on
corruption. This analysis is not sufficient evidence to demonstrate that
International Aid is the cause of the
change in HDI. This study does not correct for the inherent reciprocal
relationship between Aid and Development. Future studies could benefit from
other statistical techniques such as time lagged two stage regression analysis
utilizing instrumental variables to determine the true direction of causality.
Though these results suggest that aid is
much less effective in the presence of a corrupt government it is possible that
aid, employed correctly, can improve corruption thus paving the way for more
effective aid in the future. Additionally, this analysis examines aid in the
aggregate and does not distinguish between different types of aid or different
delivery mechanisms. Despite these complications, organizations like the World
Bank and Transparency International have frequently documented the challenges
of getting aid to the people and projects that need it most in countries with
rampant government corruption. This raises significant questions about the
whether or not aid should precede other measures of addressing public
corruption.
Table 1. Effect of Corruption on
International Aid
|
||||
DV:
Change in HDI Index
|
(1)
|
(2)
|
(3)
|
|
Total of International Aid (log)
|
0.06***
(4.97)
|
0.002
(2.02)
|
0.0008*
(0.59)
|
|
Improving Corruption
|
-0.49*
(-2.10)
|
|||
Imp. Corr * Aid (log)
|
0.004**
(2.07)
|
|||
Education
|
-.018*
(-1.02)
|
-0.015*
(-0.85)
|
||
Life Expectancy
|
-.0004
(-1.48)
|
-0.0004
(-1.50)
|
||
Trade (% GDP)
|
-2.1e-6
(-0.05)
|
-0.00001
(-0.28)
|
||
Constant
|
-.02
(-1.36)
|
0.044
(1.55)
|
0.065
(2.19)
|
|
Observations
R2
|
115
0.18
|
115
|
115
0.26
|
|
Note: OLS estimates with t-stats in parentheses. * p < 0.05, ** p < 0.01, *** p <
0.001
Source: Worldwide Governance Indicators, World Bank Development
Indicators
|
||||
Appendix A
Summary Statistics and Variable Descriptions:
Table 2. Summary
Statistics
|
|||||||
Variable
|
Label
|
Mean
|
Std.
Dev.
|
Min.
|
Max.
|
Description
|
|
International Aid (log)
|
A
|
13.34
|
2.01
|
7.86
|
17.40
|
Logarithm of total international aid 2003-2011
|
|
Education
|
B
|
0.50
|
0.16
|
0.15
|
0.77
|
Index of educational attainment (0 = Min, 1 = Max)
|
|
Life Expectancy
|
C
|
64.08
|
9.70
|
42.2
|
78.4
|
National aggregate life expectancy in years
|
|
Trade (%GDP)
|
D
|
83.24
|
44.72
|
27.06
|
307.02
|
Country level trade as a percentage of GDP
|
|
Corruption
|
E
|
-0.03
|
0.34
|
0
|
1
|
Change in Corruption index 2003-2012. (0 = Worsening,
1 = Improving)
|
|
Observations
|
115
|
||||||
Overall this is an important question, a well-designed study, and the author convincingly builds a case for why we should care about its findings. While I understood the research question, I found the identification of the IV and DV somewhat unclear, and was only able to figure out what they were after looking at the graph. After looking at the graph it becomes more clear what the IV and DV are.
ReplyDeleteIt is not clear that the blog post answered the question being asked. In looking at the final regression table it appears that none of the coefficients in the 2nd and 4th models were significant. These two models were the only ones that contained the corruption dummy variable. The conclusion reached, that corruption has a dampening effect on international aid, is certainly suggested by the second graph, but not by the significance of the coefficients for the final model, all of which are not significant.
The first graph only has a lowess line, but should also have an OLS line. Table has model #s 1,2,4, but no #3, which is never explained.
The author successfully examined linearity and made the necessary transformations. He does not rule out reverse causality, but certainly accounts for it, explains how it might affect the study, and suggests another study design that would better control for it.
In general, this was a thoughtful analysis on an interesting topic that is relevant and people should care about reading it. The title made me think you were going to write about corruption within the international aid sector and to specifically about governments. I am not sure that HDI is the best dependable variable to use as a proxy of goverment corruption. The author wrote for an audience that has statistics background. He did a great analysis on about confo
ReplyDeleteThe author tested for linearity and made the transformation but the results seemed to be not statiscally significant. I don't believe the author answer the question.
Very nice graphs, tables and analysis.
Big Picture
ReplyDeleteIs the research question clear? The research question is clear, the introduction could have touched more on previous research that points out the interaction of corruption and foreign aid in developing countries.
Independent variable: Foreign aid
Dependent variable: HDI
Is it clear why this is an important question? Yes, you make a good point in stating the importance of including corruption in your analysis.
Does the author answer the question? Yes, the author addresses the question but does not reference the lack of significance in the results. It is clear by the information given by the author that corruption is important in studying aid and hdi but does not further explain his findings in relation to that statement.
Nuts and Bolts
Lowess, OLS? There is a graph with a lowess line but the regression line is missing.
Summary, regression tables? Yes, both the summary table and the regression table are part of the post. All tables and graphs are well presented and clear.
Modelling and Inference
Linearity? Yes, the author did examine linearity and transformed variables.
Cofounds? Yes, although it is not clear why you chose to use trade as % of GDP instead of GDP per capita. Also, maybe it would have been useful to use international aid per capita as well.
Reverse causality? The author addresses the problem of reverse causality but also states that he does not control for it in this model.
Results- the interpretation is clear however there is not clear reference to the results, especially to the lack of significance of the coefficients.
Overall, great job. You did a great job in framing the problem and contextualizing it. The language throughout the post is very clear and direct.