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

 

Insert Table 2 about here 

 

First, the table throughout shows a strong conditional convergence as the initial 

GDP per capita exerts a strongly negative effect. As is to be expected, the investment 

rate also remains significant and with a considerable effect although this appears mainly 

to be the influence of a number of well-performing countries. When those observations 

are excluded, the overall investment rate is not significant (columns 7 and 8), consistent 

with the findings in Easterly (1999). As the last of the control variables, government 

consumption is only significant in one case but has a positive sign in all but one case. 

Turning to the variables of interest in this paper, odd-numbered columns report the 

results of the baseline specification while even-numbered columns add an interaction 

between foreign aid and debt. The results rather clearly show that having a debt burden 

significantly lowers the growth rate. Foreign aid, on the other hand, does not exert any 

robust effect on its own. Even though it is significant in three out of the four odd-

numbered columns, excluding the worst performing countries (column 5) in a group of 

LDCs that are already characterized by having slow or no economic development 

indicates that the beneficial effects of aid per se only pertain in truly deep crisis. 

However, when adding an interaction term in the even-numbered columns of 

Table 2, the picture changes somewhat. The interaction term between foreign aid and 

external debt is individually significant in all but one case, and aid and the interaction 

term are strongly jointly significant in all cases (cf. the bottom line of the table). 

Moreover, the interaction term is substantially larger in the sample that restricts 

countries to have at least four observations and thereby minimizes the potential problem 

 

10

---

of spurious effects when countries enter the sample. This set of estimates, which is 

arguably the most reliable, points to a rather considerable growth suppressing effect of 

debt burdens in LDCs. This effect is roughly halved if the inflow of foreign aid 

increases by 30 percentage points. For the typical country in the sample, this would 

nevertheless imply that aid would have to be tripled. As such, the table both shows 

considerable support for the theoretical conjecture in this paper as well as indicating that 

the alleviating effects should not be overstated.

4

 We return to discussing this point later. 

 

Insert Table 3 about here 

 

Table 3 instead explores whether there also is an interaction effect on the 

investment rate. To explain the investment rate, we include a baseline specification 

consisting of the logarithm to initial GDP per capita and the government consumption, 

which both were part of the baseline model in the growth regression in table 2. Further 

we expand the basic regression by including the openness to trade and the real 

investment price, which must necessarily be considered an important determinant of the 

investment rate. 

                                                 

4

 With respect to the estimates in Table 2, it is often the case that the inclusion of government 

consumption in growth specifications has a tendency to bias aid coefficients upwards, as the positive 

effects of aid on consumption are controlled for. Yet, a series of additional regressions without 

government consumption reveals that this is not a serious concern in the present sample of LDCs. In the 

full sample, the coefficient on the interaction term is somewhat reduced but remains significant at p<.10 

while the sample in columns 3-4 shows no significant difference with and without government 

consumption. As such, our conjecture in section 2 that fungibility may not be an important issue in LDCs 

receives some support. 

 

11

---

First, economic development as measured by GDP per capita shows a significant 

and positive association with the investment rate in three out of four odd-numbered 

columns. Yet, this association becomes insignificant when the analysis is limited to only 

countries with investment figures for at least four periods. The variables for openness to 

trade and government consumption both enter the regression with significant and 

positive coefficients, implying that openness affects the investment rate positively as is 

the standard conclusion (cf. Levine and Renelt, 1992), while the government 

consumption most likely picks up the fact captured in the theoretical model of this 

paper, that public investments constitute a substantial share of total investments in many 

LDCs. On the other hand, the real investment price not surprisingly affects the 

investment rate in the opposite direction, having a rather large negative and significant 

coefficient in all regressions.  

Turning to the variables of interest in this paper, the table shows that neither the 

debt burden nor inflows of foreign aid significantly affect the investment rate as they are 

insignificant in all odd-numbered columns. Adding the interaction term between foreign 

aid and the debt burden in the even-numbered columns does not imply that the two 

interesting variables turning significant, but they switch sign to becoming negatively 

insignificant. The interaction term, on the other hand, enters the regressions with 

significant positive coefficients in three out of four regressions, meaning that foreign aid 

in conjunction with large debt burdens has a positive impact on the investment rate. The 

only deviation from this conclusion is that the interaction term, like GDP, turns 

insignificant when the regression only contains the countries with investment figures for 

at least four periods. However, the inclusion of debt, aid and the interaction term are 

jointly significant in all four regressions. 

 

12

---