Working paper 07-1 Christina Bjerg, Christian Bjørnskov and Anne Holm

(

)

(

)

t

t

t

t

t

t

t

a

erd

m

ep

d

d

e

a

x

θ

θ

−

+

+

=

−

+

+

−

1

*

1

  

 

(1) 

 

This leaves the slightly absurd theoretical possibility to be tested in this paper: that 

foreign aid only has positive growth effects to the extent that it is used to alleviate 

external problems and thus does not enter the economy. In other words, it is possible 

that aid could have positive growth effects when it is used contrary to its intended 

purpose. For a developing country with no or only little external debt, foreign aid 

inflows would result fully in Dutch Disease problems and thus lead to an ambiguous 

relation with economic growth. In developing countries with high levels of external 

debt, on the other hand, foreign aid would provide much-needed foreign currency to 

meet debt payments. In the latter situation, a substantial part of the aid would therefore 

not enter the economy, leaving less severe Dutch Disease and alleviating debt problems, 

and thereby making a positive growth-aid association more likely. The specific 

hypothesis to be tested in the following is therefore: 

 

Hypothesis: The association between foreign aid inflows and economic growth 

rates depends positively on the country’s level of external debt. 

 

3. Data 

To be able to test this proposition, we employ data from different sources. The data to 

be employed in the next section all derive from the Center for Global Development 

database (CGD, 2006), the Penn World Tables Mark 6.1 (Heston et al., 2002) and the 

World Bank’s (2005) World Development Indicators database. Our dependent variable 

in the following set of Barro-type regressions is the average annual growth rate across 

 

7

---

each five-year period between 1960 and 2000 for which full data are available. To 

explain these growth rates, we employ a fairly simple baseline specification including 

the logarithm to initial GDP per capita in PPP-adjusted US dollars, the investment rate 

as percent of GDP and government consumption as percent of GDP, all from the Penn 

World Tables. In a set of investment regressions, we also include openness to trade and 

the relative investment price (capital goods prices as a ratio of the overall price level) 

from the same source. The specifications also include our variables of interest, foreign 

aid as percent of GDP, taken from the Center for Global Development database, and the 

external debt as percent of GDP, taken from World Development Indicators and 

averaged across each five-year period, as well as period fixed effects to account for 

global business cycles. In half of the regressions below, we also include an interaction 

term between aid and external debt, based on the centred values of the variables. As 

such, the size and significance of the interaction term is evaluated around the sample 

mean (cf. Brambor et al., 2006). Table 1 gives descriptive statistics of the data; all 

countries included are reported in the appendix. 

 

Insert Table 1 about here 

 

We estimate Barro-type regressions using generalized least squares with country 

fixed effects as Hausmann tests strongly favour fixed over random effects.

2

 We present 

results in two different samples. First, we use the largest possible sample for which 

                                                 

2

 Using the full sample and specification, the Hausmann test statistics of the growth regression and 

investment regression are Chi

2

 = 19.12 (p<.004), and Chi

2

 = 24.92 (p<.009), respectively. Without the 

interaction term between aid and debt, the test is marginally stronger for the investment regression and 

somewhat stronger for the growth regression. 

 

8

---

there are full data, consisting of 199 observations from 38 of the 49 countries currently 

categorized as LDCs. As noted above, we restrict our attention to LDCs as these 

countries both have large debt problems as well as very limited inflows of foreign direct 

investments. However, the possibility exists that these results may be biased as 

countries may tend to enter international datasets after periods of relatively good 

performance, which could bias coefficients on both foreign aid and external debts 

upwards.

3

 Second, by restricting the sample to those countries for which there are data 

for at least half of the period 1960-2000, an alternative sample comes to consist of 156 

observations from 27 countries. We also run separate analyses excluding either the top 

performers or bottom performers at any time in the LDC group. 

 

4. Empirical results 

Before turning to the formal empirical results, it can be informative to look at some 

simple statistics. First, the data from the 38 LDCs show the same picture as the overall 

data from developing countries, as the simple correlation between economic growth and 

foreign aid is far from being significant at -.02. On the other hand, the correlation 

between growth and the ratio of external debt to GDP is -.13 and significant at p<.10. At 

first sight, then, it would seem that the raw data indicates that outflows of interest 

payments on debt could be more important than inflows of foreign aid. Whether this is 

the case is further explored in Table 2 that reports the results of estimating growth rates 

for the LDCs. 

                                                 

3

 That this is not merely a potential problem within this sample is indicated by some simple sample 

statistics. In the first period that countries enter, the sample average growth rate is .64 percent while in the 

following period, average growth is -.71 percent and -.10 on average for all subsequent periods. 

 

9

---