World Politics. Develop a research design to assess the impact of any one of the following: (1) bilateral U.S. foreign aid, (2) multilateral foreign aid to any specific country of your choosing (3) any third world country's population policy, (4) any U.N. specialized agency program, or (5) any U.S. domestic or foreign program (at the state, local, or
national level) of your choosing.
Your discussion needs to focus specifically on bilateral “US” foreign Aid!
Generally, you need to specify your expectation clearly. Sometime your discussion is very vague and not coherent. To improve your weaknesses, you need to describe the following points precisely and constantly.
What is your research question? For instance, Does US bilateral aid increase recipient country’s GDP growth??
How do you approach to it?
How do you hypothesize your expected relationship?
How do you measure your dependent and independent variable? What is your data soruce. Why your measurement is valid?
What are your control variables if necessary?
How do you analyze data ?
How do you interpret the statistical results? direction, strength, statistical significance
What is the substantive meaning of statistical results?
What is your study’s weakness or strength?
Theory
o There is the significant literature on the determinants of aid, specifically: a few examples of which are Robert D. McKinlay and Richard Little (1978, 1979), Alfred Maizels and Machiko K. Nissanke (1984), Bruno S. Frey and Friedrich Schneider (1986), and William N. Trumbull and Howard J. Wall (1994).
o In general, this literature has found that donors' strategic interests play an important role in the allocation of aid, whereas commercial interests have not been as important.
o Furthermore, more aid is given to countries with low income, and aid relative to GDP is much higher for countries with small populations.
o The effectiveness of foreign aid remains an unresolved issue.
o Most studies focus on the impact of aid flows on Gross Domestic Product (GDP) and other macroeconomic variables such as investment or public consumption.
o The most influential aid effectiveness literature stems from Chenery’s and Strous’ (1966) two-gap model, which focuses on aid-growth and aid-savings relationships.
o But most early authors concluded that aid had no significant impact on growth, savings or investment.
o Aid was shown to increase unproductive public consumption (Mosley and others, 1992) and to fail to promote investment.
o The latter point is confirmed by Boone (1996) and Reichel (1995) who find a negative relationship between savings and aid, and point to a substitution effect.
o This result is amended by Hadjimichael (1995), who notes that the relationship between aid and domestic savings is negative in most countries, but positive for good adjusters.
o The latter point is confirmed by Burnside and Dollar (2000)
o So, This body of literature has produced inconclusive results.
Which theory do you want to test? Determinants or Effectiveness?
If you are interested in studying effectiveness of US bilateral foreign Aid, your discussion could focus only on effectiveness. Based on your argument, I am not convinced why you want to study the effectiveness. Moreover, I am wondering what the effectiveness means?
Statement of Research Goals
o I propose a two-pronged approach to measuring the overall impact of U.S. bilateral aid
(1) Measure aid effectiveness by GDP growth in recipient.
(2) At times, it is possible to gather information about how U.S. bilateral aid is earmarked for certain spending purposes. Specifically, I am interested in whether or not U.S. aid to LDCs reduces poverty what does it mean “Poverty”? You need to define the term .
· Where possible, I follow Boone (1996), Masud and Yontcheva (2005) and others in testing also for the impact of U.S. bilateral aid on Human Development Indicators (HDIs).good!.
· As indicated by Boone (1996), human development indicators respond quickly to improved services and therefore can be considered “flash indicators of improvement in the conditions of the poor.” Okay so what do you want to do?
· I focus on infant mortality, life expectancy, literacy and income poverty levels, because health and education indicators are more concrete measures for analyzing poverty than macroeconomic indicators. (Masud and Yontcheva, 2005) Good! So you want to use this instead of HDIs?
Hypotheseis
H(1) = US foreign aid to LDCs increases GDP growth in those countries receiving aid.
H(2) = US foreign aid to LDCs decreases poverty in those countries receiving aid.
H1 and H2 are saying that US foreign aid has positive impact on higher GDP growth and lower Poverty. Why did you expect such relationship? Are you suggesting if a country receive US aid, her GDP Growth rate increases or is higher than one of non-received countries ? It sounds ad-hoc, too.
Research Design
Unit of Analysis
o US foreign Aid to 120 LDCs in a given year
Here your unit of analysis or observation is not aid itself but aid flow from the United States to 120 individual countries between 1970 and 1999.
Stimson (1985) notes that the in political time-series it is “endemic” that we are typically confined to annual data as finer levels of analysis are scarcely available.
This is true, but what can you do?
Definition of Variables and Measures
Dependent variable (1): GDP in per capita in 120 LDCs, annually (constant in 1995 U.S. dollars); à NOTE: GDP per capita is not equal GDP Growth, which means change.
To measure GDP growth, you need to state that GDP per capita yarer t – GDP per Capita year t-1
Dependent Variable (2): HDI for the 120 LDCs for which reliable aid data is available for 1970-1999. But you said above, “I focus on infant mortality, life expectancy, literacy and income poverty levels”. Do you want to disaggregate HDI to each factors or want to use HDI as an index? Confusing???
Technically, previous US aid xt-1 affects GDP growth t ad HDI t.
Independent variable: U.S. bilateral aid to 120 LDCs from 1970 to 1999 measured in 1,000 U.S. dollars, annually. OK, but it seems that you want to measure the presence or absence of US aid, isn’t it??
Dependent variables: GDP in per capita in LDCs, annually (constant in 1995 U.S. dollars); HDI for the 120 states for which reliable aid data is available for 1970-1999.
Unit of Analysis
o
Stimson (1985) notes that the in political time-series it is “endemic” that we are typically confined to annual data as finer levels of analysis are scarcely available.
Data Sources
o GDP per capita from World Development Indicators, World Bank OK!
o Statements of bilateral aid from the Organization for Economic Cooperation and Development (OECD) OK, but USAID also has the US bilateral aid data.
o A review HDI available from the United Nations Development Programme (UNDP) and commentary regarding data availability from Kemmer (2004). Ok!
o Infant Mortality from World Development Indicators, World Bank Ok, but what for you need this data? You did not hypothesize any relationship bilateral AID flow and infant mortality, right??
Research Design
o quasi-experimental design
o I will employ a multiple time series design, heretofore referred to as time-series cross-section (TSCS) what does it mean? Why do you think this design is appropriate? Time series and cross section? Beause of 120 countries between 1970 and 1999.right?
o Masud and Yontcheva (2005) control for the level of development represented by per capita GDP, the poverty headcount, the level of rural development, and female illiteracy.These are your control variables? How do you measure them? You might need to mention.
o They also assess the impact of government efforts in reducing infant mortality as represented by the per capita health expenditure. So what do you mean? Do they control for government effects such as democracy or autocracy? For democracy and poor, see Michael Ross (2006) in American Journal of Political Science.
What are the strengths and weaknesses of quasi-experimental design?
o Strengths:
§ Although we cannot control for all extraneous variables, this is really the best we can do if we hope to make any causal inferences about U.S. bilateral aid.
o Weaknesses:
§ We cannot meet the random assignment condition that is necessary for true experimental design. U.S. does not randomly select countries who receive aid.
§ We are not conducting the experiment in a controlled setting.
à I do not think that you need this discussion because you specifically discuss your research design below.
What are the strengths and weaknesses of TSCS? This is a good discussion.
o TSCS is most appropriate for these situations where we have systematic comparisons of a limited number of units across time, but we do not have random assignment to a control group.
o In TSCS, analysts typically observe covariation assumed to be produced by unobserved causal processes operating at some time before the data in question were gathered (Stimson, 1985).
Internal Validity
o Major threat to internal validity is history. Something else may account for a change in our dependent variables. For example, there could have been some sort of economic rebound, aid from NGOs during the same time period, end of some conflict or a corrected inefficiency, etc. à can we control it?
o We’re also dealing with selection bias in that U.S. bilateral aid is not randomly assigned to recipient countries. It is the product of a multi-layered domestic and international process.??? The recipient countries may be already rich or high HID??
o Selection effect possible in U.S. giving aid to countries more likely to do well, less to those not likely do well.Yes
o Three main arguments have been proposed in the literature for what most encumbers studies of bilateral aid (Masud and Yontcheva, 2005):
(1) aid is misallocated – donors give aid for strategic reasons and to the “wrong” recipients
o In fact, Burnside and Diamond (2000) find that self-interest is the primary motivator when donor states make decisions about aid allocation, but you want to estimate the effect of aid on GDP growth
(2) aid is misused – recipient governments pursuer non-development agendas OK
(3) GDP growth is generally not the right measure of aid effectiveness Right! But so you want to use HDI, isn’t it??
External Validity
o we cannot necessarily generalize these results to all. Why? The results will be generalizable for the impact of US bilateral Aid flow on GDP growth and Poverty. You have enough sample with Time series and cross-sectional design. (120 countries x 30 years=3600)
Method of Analysis
o First, I would conduct an independent samples differences of means test, using a t-test. What for? Explain!Then, I would compare the mean values of the dependent variables in years where no aid was given t0 and to their mean values in years receiving aid, (t0 and later.) What does this discussion mean? What do you want to know about this procedure? Do you want to check there is normal distribution or skewed distribution? If you find skewed distribution, what could you do?
You wanted to code your independent variable as dichotomy.
o In using TSCS, we deal in employ time-series regression analysis with s associated with Box and Jenkins. .
o I would use a regression model, regressing yt against year (x1=1, first year of series, second year in series, etc.), against a dummy variable (x2 = 0 for non-aid years, and x2= 1 for aid years) to account for differences in intercepts, and also against an interaction term (x3= x1*x2) to allow for different slopes. I do not understand what this interaction term means.
Can you write equation?
Here your argument has been changed. You should have two dependent variables instead of one. Also your measurement of independent variable are changed.
Yt = GDP growth or HDI or Infant Mortality
X1t = ???
X2t= AID (Yes or no) in previous year?
X3t= control variable?
X4t= control variable?
X1t affects or causes Y1t in year 1 for a given country
f X1t2 affects or causes Y1t2 in year 2 for a given country
Box and Jenkins (1970), Christensen (2001) and many others suggest that the most appropriate statistical test here is the Bayesian moving average model. This approach requires many data points, but should help us determine if the pattern of measures taken in years of aid differ from measures taken in years where no aid was given.
What are potential problems?
o Autocorrelation: In regression analysis using time series data, autocorrelation of the residuals ("error terms", in econometrics) is a problem. Yes!
§ Autocorrelation violates the OLS assumption that the error terms are uncorrelated. While it does not bias the OLS coefficient estimates, the standard errors tend to be underestimated (and the t-scores overestimated).yes
§ The traditional test for the presence of first-order autocorrelation is the Durbin–Watson statistic Yes
o Heteroscedasticity: Heteroscedasticity is a violation of the assumption that the error term has a constant variance. Yes
o does not cause OLS coefficient estimates to be biased. However, the variance (and, thus, standard errors) of the coefficients tends to be underestimated, inflating t-scores and sometimes making insignificant variables appear to be statistically significant
Special Problems in Analyzing Bilateral Aid:
o What about amount of aid? We can’t assume the aid amount is constant year-to-year. Yes so you should not dichotomize
§ Zuk and Thompson (1982) in particular discuss the problem of having to contend with a study where units may remain the same, but the amount of dollar flowing in to them can easily vary by significant amounts from year to year.
§ They propose standardizing the data, by dividing each cross-sectional unit by a constant (in their example, 100) so that there will be an arbitrary base. Could be, but not so important if your measurement is consistent.
§ But as Stimson (1985) notes, this creates artificial parameter estimates. Okay.
o Need to investigate whether or not the donor earmarked aid for a specific purpose. What for?
Testing
I will reject the null hypotheses of no impact due to U.S. bilateral aid if a particular coefficient has a two-tailed significance less than or equal to .05.
So What does it have substantively meaning? Is it statistically significant? Positive or negative or strong or weak?
Substantive Importance
I would judge a 5% cumulative reduction in infant mortality rates due to U.S. bilateral aid as being a substantively important program impact. Why only infant mortality rates? How about GDP growth and HDI??
1