January 2016

The Impact of External Debt on Economic Growth:

Empirical Evidence from Highly Indebted Poor Countries*

by

Abu Siddique, Business School, The University of Western AustraliaE A Selvanathan, Griffith Business School, Griffith University, and Saroja Selvanathan**, Griffith Business School, Griffith University

Abstract During the 1970’s and 1980’s, the external debt levels of poor countries rose to a level constituting a ‘debt crisis.’ The main source of external debt was the surplus revenue generated by significant increases in the price of oil during the 1970s. Unfortunately, many of the countries failed to use the external debt wisely and prudently. When the revenue from oil sales started to decline due to low oil prices during the 1980s, heavily indebted poor countries (HIPCs) experienced difficulty servicing the debt. Using HIPCs data, this paper analyses the extent to which the external debt burden impacts on a country’s GDP.

Key words: External debt, Economic growth, highly indebted poor countries, Debt relief.

JEL Codes: O10, F34, H12, H63, O47, O55

* The authors would like to acknowledge and thank the four anonymous reviewers and the editor of this journal for their constructive comments on an earlier version of the paper.

** Corresponding author: Professor Saroja Selvanathan, Economics and Business Statistics Discipline, Griffith Business School, Griffith University, Nathan campus, 170 Kessels Road, Nathan, Queensland, 4111, Australia; email: s.selvanathan@griffith.edu.au.

The Impact of External Debt on Economic Growth:

Empirical Evidence from Highly Indebted Poor Countries

1. Introduction

External debt is an important source of finance mainly used to supplement the domestic sources

of funds for supporting development and other needs of a country. Usually, external debt is

incurred by a country which suffers from shortages of domestic savings and foreign exchange

needed to achieve its developmental and other national objectives. However, if the external debt

is not used in income-generating and productive activities, the ability of a debtor nation to repay

the debt is significantly reduced. It is often argued that this excessive debt constitutes an obstacle

to sustainable economic growth and poverty reduction (Maghyereh and Hashemite, 2003; and

Berensmann, 2004). Over the 1970’s and 1980’s, the external debt levels of ‘highly indebted

poor countries’ (HIPC)1 rose to a level constituting a ‘debt crisis.’ The bulk of this debt was

made up of public and publicly guaranteed debt (PPG). The main source of this supply of

external debt was the emergence of the Eurodollar market resulting from the surplus revenue

generated by the Organisation of Petroleum Exporting Countries (OPEC) through significant

increases in the price of oil between 1973 and 1979. Cheap ‘petrodollars’ were recycled to the

countries which needed external debt. Unfortunately, many of the countries failed to use the

external debt wisely and prudently. A number of interrelated factors contributed to the rise in

external debt including macroeconomic policy, increases in the price of a number of primary

commodities encouraging countries to borrow, low real interest rates and a favourable world

environment. Unfortunately, the favourable conditions were short-lived and, when they did

change over the 1980’s, heavily indebted countries experienced difficulty in servicing the debt

(for details, see Barro, 1989; Altvater, 1991; Abbott, 1993; Barro and Lee, 1994; Siddique, 1996;

Abrego and Ross, 2001; and Clements et al, 2003).

1 HIPC countries have 3 main common characteristics: (a) they incurred heavy debt mainly in the 1970s but the symptoms of HIPCs emerged in 1980s; (b) debt ratio of these countries is much higher than other low income or developing countries; and (c) they are poor countries with a lower economic growth (Birdsall et al., 2002).

1

In 1996, the International Monetary Fund (IMF) launched the HIPC initiative in an

attempt to reduce the external debt burden of low-income countries to sustainable levels in a

reasonably short period of time2. The HIPC initiative has generated a lot of attention and has

been hailed by many as a significant event, promising economic opportunities for debt ridden

poor countries. How much benefit did this ‘initiative’ bring for the HIPC? Literature examining

the relationship between reductions in external debt and economic growth gives mixed signals to

the policy makers in both the developed and the developing countries. It suggests that while debt

can have a positive short term economic growth impact up to a certain ‘threshold’ point (useful

for smoothing out business cycle fluctuations in unemployment and inflation) and it can be a

valuable source of finance for development/ investment in HIPCs that otherwise could not self-

finance, debt can also have a negative economic growth impact in the long term and beyond the

‘threshold’ level. Consequently, for HIPCs which have passed this ‘threshold’, the best way to

ensure the effectiveness of appropriate lending and economic reforms is to combine them with

debt relief.

Several research papers that analyses the impact of external debt on the economic well-

being of either a single or a group of countries have been published, starting from the early

1980’s (for example, see Adams et.al, 1983; Kwack, 1983; Filatove and MaHiona, 1985) and to-

date (for example, see Gomez-Puig and Sosvilla-Rivero, 2015; Hallett and Oliva, 2015; and

Timo and Florian, 2015). The main objective of this paper is to examine the influence of a

change in external debt on economic growth in the HIPC countries over the period 1970 to 2007,

using recent developments in time series and cross-sectional analysis. Standard growth

accounting process done by decomposing the sources of economic growth will be employed for

2 In order to qualify for HIPC Initiative assistance, a country must meet the following four conditions: “1. Be eligible to borrow from the World Bank’s International Development Agency, which provides interest-free loans and grants to the world’s poorest countries, and from the IMF’s Poverty Reduction and Growth Trust, which provides loans to low-income countries at subsidized rates; 2. Face an unsustainable debt burden that cannot be addressed through traditional debt relief mechanisms;3. Have established a track record of reform and sound policies through IMF- and World Bank-supported programs; and4. Have developed a Poverty Reduction Strategy Paper (PRSP) through a broad-based participatory process in the country.” (IMF, 2014).

2

this purpose. In addition, the paper will analyse the extent to which the external debt held by

heavily indebted poor countries has impacted on their economic growth.

2. The Relationship between External Debt and Economic Growth: A Brief Review

In this section, we briefly review the literature examining the impact of external debt on

economic growth based on a number of selected studies.

Using data covering 1969-1998, Pattillo et al. (2002) analyse the effect that debt burden

has on 93 developing economies. They have included, in their analysis, a set of variables

common in the growth literature, namely that of trade openness, schooling, population and

government budget. In summary, they found that for a country with average indebtedness, a

doubling of the debt ratio would reduce annual per capita growth by between a half and a full

percentage point. Furthermore, they found that the average impact of debt only becomes negative

at debt ratios above 160-170 percent of exports or 35-40 percent of Gross Domestic Product

(GDP), and that the marginal impact of debt only starts becoming negative at about half of these

levels.

Clements et al. (2003) seek to analyse the channels through which external debt impacts

upon economic growth in low-income countries. They estimated a reduced form growth equation

for 55 low-income countries using data from 1970 to 1999. Their results support the debt

overhang hypothesis (which is?) and they estimated a threshold level of debt to exports of 100-

105 percent; and a threshold level of debt to GDP of 20-25 percent.

Contrary to the above findings, Jayaraman and Lau (2009) found that higher debt levels

can promote higher economic growth. Their study involves six Pacific island countries, covering

the period 1988-2004, and is based on regressing external debt stock, exports and the budget

deficit (all as a percentage of GDP) against GDP. They found that a 1 percent increase in the

external debt stock leads to a 0.25 percent increase in national output. They also tested for

causality and found that, whilst there is no Granger causality relationship between real gross

3

domestic product (RGDP) and external debt in the long-run, there is a significant causal

relationship running from external debt to GDP in the short-run.

Hameed et al. (2008) analysed the relationship between external debt and economic

growth in Pakistan, using data for 1970-2003. They identified a long-run relationship between

the external debt and economic growth; and concluded that debt service affects GDP negatively,

most likely through its adverse impacts on capital and labour productivity. They also found

short-run and long-run negative causality that runs from debt service to GDP.

Adegbite et al. (2008) investigated the impact that Nigeria’s huge external debt stock had

on its economic growth between 1975 and 2005. They used a ‘Solow-type’ neoclassical growth

model to regress the ratio of external debt to GDP (along with several other macroeconomic and

external sector exogenous variables) against the annual GDP growth rate. Their investigation of

the debt overhang theory for Nigeria revealed that external debt contributes positively to growth

up to a certain point, after which its contribution becomes negative. They also investigated the

‘crowding out’ effect of debt servicing by regressing debt service requirements against private

investment and found that Nigeria’s large debt burden did indeed ‘crowd out’ private investment.

Fonchamnyo (2009) studied the effect of economic and social performance in 60 low-

income countries to assess the relative effectiveness of the HIPC Initiative. He divided the 60

low-income countries into four groups, based on their 2005 HIPC status; non HIPCs; pre-

decision point; decision point; and completion point HIPCs. His hypothesis was that these

countries included in the HIPC Initiative will show better improvement in economic and social

development than those countries not included. From the estimation results of an investment

function and an economic growth function, the study concludes that investment and growth have

improved in HIPCs since the institution of the HIPC Initiative; and, there is also evidence that

health care and education enrolment experienced some improvement in countries that had

reached the completion point of the HIPC Initiative. (Not shown)

Fosu (1999) studied the effect of external debt on the growth of 35 countries in sub-

Saharan Africa using World Bank data for 1980-1990. By regressing GDP growth on the growth

4

rates of labour, capital, exports, and external debt, Fosu (1999) found that net outstanding debt

has a negative effect on economic growth (for given levels of production inputs). Furthermore,

he found that growth across these sub-Saharan African nations would have been 50 percent

higher during the period of study in the absence of the debt burden; and little evidence of a

negative correlation between external debt and investment levels.

Faini and de Melo (1990) assess the success of adjustment packages to developing

countries supported by loans from the World Bank and IMF; focusing on a series of

microeconomic reforms to assure supply-side improvement whilst simultaneously pursuing sharp

real exchange rate depreciation. The authors found that high external debt burdens, in

conjunction with macroeconomic instability, impede investment in developing countries. They

argue that, for such adjustment packages to result in the levels of investment necessary for the

packages to succeed, appropriate relief of external debt is required.

Froot (1989) compared different market-based debt reduction schemes, and argued that

the optimal approach to debt relief is a package that is part debt forgiveness, and part new

lending. In particular, Froot (1989) found that debtor nations that finance buybacks using current

resources can impede incentives for new investment, and can, therefore, prolong the debt relief

process.

Fry (1989) examined the effect of foreign debt accumulation on the balance of the current

account, using data from 28 countries identified to be heavily indebted to the World Bank, in

1986. He argued that, as long as an increase in foreign debt increases investment by less than it

increases saving or reduces investment by more than it reduces saving, then the current account

will enter a state of equilibrium with a maintainable ratio of foreign debt to gross national

product. In particular, Fry identified public and publically guaranteed debt as reducing saving by

more than it reduces investment, hence worsening the current account deficit over time.

Siddique (1996) analysed the external debt problem facing 32 sub-Saharan African

nations over the period 1971 to 1990, and found the following three major interdependent factors

contributing to the accumulation of external debt, namely, the trade policy, the macroeconomic

5

policy and the external and global shocks. The study also identified cases in which

macroeconomic policy designed to achieve high economic growth enhanced the ability of debt-

stricken low-income countries to meet their debt obligations.

3. Emergence of the Debt Problem

3.1 Factors Contributing to the Emergence of the HIPC Countries

The emergence of unsustainable debt in the HIPC countries can be analysed from both the

demand and supply side. From the demand side, the group of countries which are now classified

as HIPCs needed external debt to meet their development and other needs. Most of these

countries were poor with relatively lower economic growth and lower per capita income. Hence,

national rates of savings were also very low with domestic savings being insufficient to finance

their developmental and national goals. Moreover, as most of these countries were dependent on

the exports of primary commodities, their export earnings were not enough to finance import

bills as they mostly imported capital intensive goods which were relatively more expensive to

the earnings from primary products. Hence, there arose the need for external borrowing.

Fortunately (or, unfortunately) due to the significant increase in the price of oil between

1973 and 1979, the foreign reserves of oil exporting countries dramatically increased with most

of these reserves deposited mainly with European banks. Thus, there arose a market for external

debt or borrowing from overseas sources.

During the 1970s, most of the governments of developing and poor countries heavily

borrowed money primarily to finance their industrial and infrastructure development. There was

a popular belief amongst many of the developing nations that their countries’ economic success

depended on industrial development which needed protection from overseas competition during

the initial stages of development. Thus, industrial development was pursued with the aid of the

import substitution industrialisation strategy. Unfortunately, many of these countries, especially

the sub-Saharan African countries failed to invest their borrowed funds in income generating

activities and, hence, failed to enhance their ability to repay their accumulated debt.

6

Based on the results reported in Siddique (1996), which analysed the external debt

problems facing 32 sub-Saharan African nations over the period 1971 to 1990, the following

three major interdependent factors were found to be contributing to the accumulation of external

debt.

The first is the trade policy. It is argued that governments seeking to address burgeoning

external debt should pursue trade policies that would result in significantly large export earnings

to meet additional debt obligations or to reduce the total stock of external debt in the long term.

Otherwise, the trade policy is considered inappropriate. Siddique (1996) found that

approximately 21 percent of growth in the stock of external debt among the low-income

countries in question was due to inappropriate trade policy during the period 1971 to 1979.

The second is macroeconomic policy. Unsustainable expansionary monetary policy can

result in a chronic current account deficit and fiscal imbalance leading to a build-up of external

debt. Furthermore, Krueger (1987) identified unrealistic macroeconomic policy as having the

potential to induce capital flight, forcing countries to engage in further borrowing not just to

meet existing debt obligations, but also to offset the impact of capital flight. Siddique (1996)

identified a number of countries, during the period 1971 to 1979, for which poor macroeconomic

policy was almost solely responsible for growth in net debt, as well as identifying cases in which

macroeconomic policy designed to achieve high economic growth enhanced the ability of debt-

stricken low-income countries to meet their debt obligations.

The third is external and global shocks. This accounts for the contribution to external

debt from factors beyond the direct control of policymakers in low-income countries, such as

real interest rates, the onset of anti-inflationary monetary policy resulting in recession, and other

global economic conditions.

Giersch (1985) identified the following three factors which contributed to the emergence

of HIPCs:

7

The 1973 oil price increase led to an increase in import bills of the indebted oil importing

countries which, in turn, led to a balance of payment deficit, causing the need for adjustment

of the production and consumption structures.

Interest rates remained relatively lower in the 1960s and the 1970s which prompted the

developing countries to borrow more than they could afford once interest rates went up.

The second oil shock in 1979 led to global recession in the early 1980s. Export incomes of

the indebted nations shrank considerably resulting in an increase in the need for more

borrowing for oil importing countries.

3.2 Trends and Patterns of the External Debt of the HIPCs

To analyse the movements in the overall trade and debt of the heavily indebted poor countries

(HIPCs), the data for the 40 HIPCs for the period 1970-2007 were obtained from the World

Development Indicators, World Bank. All the variables, except population, are in current $US.

For selected years, in columns 2-5 of Table 1 we present the aggregate total of Public and Public

Guaranteed Debt (PPG Debt), Service Debt, Other Debt and the Total Debt.3 As can be seen, a

large share of the total debt is made up of the PPG debt. The data shows strong growth in total

debt from 1974 onwards, peaking in 1998 and then continuing on a declining path but with a

slight increase in 2007. The last two columns of the table present the Total Merchandised

Exports (TME) and the Gross Domestic Product (GDP) aggregated over the 40 HIPCs. As can

be seen, there is significant improvement in the GDP, Merchandise Exports, PPG debt and total

debt of the HIPCs during the last four decades.

Table 1 Debt, Output and Export Statistics of HIPCs, 1970-2007, selected years ($US millions, current)

3 Consistent time series data for all 40 countries are available only up to 2007. The data period does not have any impact on the general applicability of the results and their policy implications presented in the paper.

8

YearPPG Debt Service Debt Other Debt Total Debt

Total Merchandise

Exports (TME)

Gross Domestic

Product (GDP) (1) (2) (3) (4) (5) (6) (7)1970 5403.1 240.4 558.4 6201.9 5971.8 26259.41974 12338.4 606.0 1585.0 14529.3 10909.3 44846.81978 27521.0 1568.2 7630.7 36719.9 13458.7 73668.81982 49704.4 3622.7 13293.7 66620.9 15787.0 96816.11986 80593.8 4220.2 17350.2 102164.2 16291.0 115507.41990 111657.5 4734.7 24296.1 140688.3 20133.0 133790.11994 128268.7 3870.9 30443.6 162583.2 19199.4 105721.41998 130308.8 4253.1 31034.5 165596.5 25443.0 141116.22002 124355.9 3802.3 25515.4 153673.7 29010.6 149686.82006 92973.6 3123.1 23890.9 119987.6 66584.9 265230.02007 95704.6 3879.9 24403.0 123987.5 75522.1 311411.7

To analyse the depth of the debt problem, we convert the data for the debt variables listed in

Table 1 in the form of relative measures with respect to GDP and merchandise exports. In

columns 2-3 of Table 2, we provide the ratios of PPG debt and total debt to GDP; and columns

4-5 of the table present the PPG debt and total debt to exports for selected years between 1970

and 2007. As can be seen, the ratios of the PPG debt and total debt to GDP (columns 2-3) and

PPG debt and total debt to exports (columns 4-5), all peaked in the mid-1990s, before declining

to close to their 1970 levels in 2007. Clearly, it appears that some progress has been made by the

HIPCs in reducing their debt level as a percentage of GDP over the last century.

Table 2 External Debt to GDP and Exports Ratios of HIPCs, 1970-2007, selected years

YearPPG Debt/

GDPTotal

Debt/GDPPPG

Debt/ExportsTotal

Debt/Exports (1) (2) (3) (4) (5)1970 0.206 0.236 0.905 1.0391974 0.275 0.324 1.131 1.3321978 0.374 0.498 2.045 2.7281982 0.513 0.688 3.148 4.2201986 0.698 0.884 4.947 6.2711990 0.835 1.052 5.546 6.9881994 1.213 1.538 6.681 8.4681998 0.923 1.173 5.122 6.5092002 0.831 1.027 4.287 5.2972006 0.351 0.452 1.396 1.8022007 0.307 0.398 1.267 1.642

In Table 3, we present the average annual growth rate of aggregated PPG debt, total debt,

GDP and exports in the 40 HIPC counties during four subsample periods and over the whole 9

sample period. During the sub-period 1971-1979, PPG debt, total debt, GDP and exports, all had

a positive double-digit growth; for the sub-period 1980-1989, the PPG and total debts grew at a

positive rate but below that of the 1971-1979 level, and the GDP and exports had a positive but

much smaller growth rate. For the sub-period, 1990-1999, the debts, GDP and exports, all had

experienced a lower but positive growth ratio. However, for the sub-period of 2000-2007, the

growth rates for the debts were negative but the growth rates of GDP and exports had recovered

and were almost in the double-digit level of the 1970s.

Table 3 Average Annual Growth Rates of PPG Debt, Total Debt, GDP and Exports in HIPCs, 1971-2007

Period PPG Debt Total Debt GDPMerchandise

Exports (1) (2) (3) (4) (5)1971-1979 22.15 24.15 13.49 12.701980-1989 12.06 11.55 4.85 1.951990-1999 2.22 2.35 0.49 3.042000-2007 -2.36 -2.47 11.66 15.39

1971-2007 8.74 9.10 7.25 7.76

4. Modelling the Relationship between Economic Performance and External Debt

The discussions in Sections 2 and 3 suggest that GDP is affected by variables such as capital

formation (CF), debt (DB), trade (TR) and population (P). In the remainder of the paper, we

consider the following variables for our analysis: (1) GDP per capita (GDP); (2) gross capital

formation per unit of GDP (CF); (3) total debt per unit of GDP (DB); (4) total trade per unit of

GDP (TR); and (5) population (P)4. The logic behind the selection of these variables is as

follows:

Capital as a share of GDP has a positive short and long term impact on GDP as a critical

input into the production process. Some industries, such as agriculture and manufacturing

can be quite labour intensive and as such, require proportionally less capital than other more

sophisticated technology and service-driven sectors. To be most effective, the local

workforce needs to be sufficiently educated and skilled (human capital) to properly utilise the

capital in more sophisticated sectors. Otherwise, the sectors may never become efficient and 4 We also included Education as a variable in our analysis, however, due to reasons given in footnote 4, Education was dropped from the analysis.

10

other more traditional sectors, such as agriculture, will have had resources diverted from

them needlessly;

Debt can have a positive impact on growth by supplementing insufficient local resources and

investment in profitable infrastructure, but only in the short term and up to a certain

“threshold” point, after which it affects growth negatively, driven by the lower efficiency of

investment as debt increases, as well as capital and labour productivity effects, and

“crowding out” effects;

Merchandise trade as a percentage of GDP has a positive relationship with GDP by

generating export revenues. This is because the products that can be exported internationally

(without artificial support) are, by definition, the products that the country is most suited to

produce (able to compete not just with international producers, but also with international

transport costs). Consequently, resources should be devoted to its production. For developing

countries, this is often agricultural or labour-intensive manufacturing produce;

Population has a long run positive impact on GDP, possibly due to its workforce or human

capital impacts.

Test for stationarity and cointegration of the model variables

Before estimation, we investigate the time series properties of each time series variable. We test

for the existence of a unit root in each panel data time series variable. We present here, the

results from five such panel unit root tests known as LLC (Levin-Lin-Chu) test , IPS (Im-

Pesaran-Shin) test, ADF (Augmented Dickey-Fuller) test, PP (Phillips-Perron) test and Hadri

test, developed by Levin et al (2002); Im et al (2003); Dickey and Fuller (1981); Phillips and

Perron (1988); and Hadri (2000), respectively.

The null hypothesis of all these tests is that the panel series has a unit root (or that the

time series is non-stationary), except for the Hadri test which tests the opposite hypothesis of the

series and has no unit roots (or the time series is stationary). The unit root test results are

presented in Table 4. As can be seen, the results indicate that capital formation and trade

11

variables are stationary in level form (that is, I(0)), whilst all the remaining variables are

stationary in their first differences, that is, I(1). In addition, we also use the Kao Residual co-

integration test to test for cointegration between GDP, capital formation, debt, trade and

population. The value of the test statistic to test the null hypothesis of no co-integration is -1.96

with a p-value of 0.0253, indicating that there is some support for co-integration between our

variables of interest. These results show that an Auto Regressive Distributed Lags (ARDL)

approach would be the most suitable to model the relationship between GDP and capital

formation, debt, trade and population.

12

Table 4 Panel unit root and cointegration test results

Individual EffectsGDP 5.250 (1.000) 3.196 (1.000) 68.965 (0.806) 52.346 (0.993) 14.654 (0.000) -Capital Formation -4.376 (0.000) -4.893 (0.000)  165.547 (0.000) 143.550 (0.000) 9.337 (0.000) I(0)Debt -1.707 (0.044) -0.372 (0.355) 72.866 (0.701) 58.800 (0.964) 9.505 (0.000) -Trade -1.908 (0.028) -2.367 (0.009)  123.291 (0.001) 119.453 (0.002) 12.889 (0.000) I(0)Population 8.759 (1.000) 21.705 (1.000) 69.619 (0.623) 7.771 (1.000) 24.479 (0.000) -Individual Effects + TrendGDP 4.496 (1.000) 2.648 (0.996) 70.232 (0.774) 36.920 (1.000) 8.065 (0.000) -Capital Formation -5.606 (0.000) -4.615 (0.000)  157.729 (0.000)  137.432 (0.000) 9.586 (0.000) I(0)Debt 8.601 (1.000) 9.824 (1.000) 32.036 (1.000) 15.303 (1.000) 10.245 (0.000) -Trade -2.745 (0.003) -2.681 (0.004)  131.994 (0.000)  139.284 (0.000) 9.097 (0.000) I(0)Population 5.559 (1.000) 12.868 (1.000) 154.118 (0.000) 28.564 (1.000) 19.195 (0.000) -

Individual EffectsGDP -19.023 (0.000) -20.094 (0.000) 539.616 (0.000) 549.998 (0.000) 1.413 (0.079) I(1)Capital Formation -33.096 (0.000) -31.743 (0.000)  876.490 (0.000)  1008.72 (0.000) -0.213 (0.584) -Debt -22.533 (0.000) -21.676 (0.000)  599.549 (0.000)  588.157 (0.000) 0.659 (0.255) I(1)Trade -31.008 (0.000) -31.172 (0.000)  855.570 (0.000) 974.278 (0.000) 3.247 (0.001) -Population 3.800 (1.000) 1.383 (0.917) 229.581 (0.000) 335.473 (0.000) 16.798 (0.000) I(1)Individual Effects + TrendGDP -17.206 (0.000) -17.015 (0.000) 419.852 (0.000) 442.905 (0.000) 9.829 (0.000) I(1)Capital Formation -26.041 (0.000) -22.947 (0.000) 795.185 (0.000) 2485.330 (0.000) 2.891 (0.002) -Debt -20.929 (0.000) -20.374 (0.000)  547.499 (0.000)  813.104 (0.000) 3.121 (0.001) I(1)Trade -26.202 (0.000) -27.212 (0.000) 729.441 (0.000) 2578.390 (0.000) 12.157 (0.000) -Population 8.297 (1.000) -0.778 (0.218)  214.845 (0.000)  391.952 (0.000) 5.221 (0.000) I(1)

Data-based value of the test statistic (p-value in parenthesis)Conclusion

LLC IPS ADF PP HADRI

B. First Differences

A. Levels

Modelling the Relationship between Economic Performance and External Debt

To model the relationship between external debt and the economic performance of the HIPCs,

we use the stationarity and cointegration test results presented in Table 4. The results of the table

suggests that a dynamic model such as an Auto Regressive Distributed Lag (ARDL) model

would be most suitable (see, Pesaran and Shin, 1999; and Pesaran et al, 2001) for our analysis, as

some of the variables are I(0) and others are I(1). We can write an ARDL (p,q,q,…,q) dynamic

panel specification for country i and period t of the form

t=1,2, …, T and i =1,2, …, N, (1)

There exists a cointegrating relationship between the variables, assuming a long-run relationship

of the form:

13

t=1,2, …, T and i =1,2, …, N, (2)

where T is the number of time periods; N is the number of countries; Xit is a (k1) vector of

explanatory variables for country i; i, is the country-specific fixed effect; ij’s are scalars and ij

are (k1) vectors of parameters of the model to be estimated. In addition, if the variables in

equation (2) are (1) and cointegrated, then the error term is a stationary process for each country

equation. This means that an error correction model can be formed in which short-term dynamics

can be combined with an error correction term to take care of the deviation from the long-run

equilibrium. Therefore, we can parameterise equation (2) in the form of an error correction (EC)

model given by:

(3)

where j=1,2,…,p-1; and

j=0,1,…,q.

The parameter i is the error correcting speed of adjustment term. If i =0, then there

would be no evidence for a long-run relationship. We would expect the estimated value of i to

be significantly negative as the variables are co-integrated; implying that the variables should

show a return to long-run equilibrium. The θi coefficients measure the long-run relationship

between the variables and, and measure the short-run relationship between the variables.

There are three approaches suggested in the literature (see Pesaran and Smith, 1995; and

Pesaran et al, 1997 and 1999) for the estimation of dynamic heterogeneous panel equation of the

form of equation (3) when both T and N are large.

The first approach is the ‘dynamic fixed effects’ (DFE) estimation where the time series

data for each country are pooled and only the intercept coefficients are allowed to vary across

countries and the speed of adjustment coefficient and the slope (short-run) coefficients are

treated as equal across countries. It has been shown that if the slope coefficients are in fact not

identical, then such DFE estimation will produce inconsistent estimates and potentially

14

misleading results. DFE estimation is also subject to a simultaneous equation bias due to the

endogeneity between the error term and the lagged dependent variable (see Baltagi et al, 2000).

The second is the ‘mean group’ (MG) estimation method proposed by Pesaran and Smith

(1995) with the aim of resolving the bias due to heterogeneous slopes in dynamic panels. Under

this approach, the model will be estimated separately for each country and a simple unweighted

arithmetic average of the coefficients will be calculated as the final (MG) estimates. MG

estimation allows for all intercept and slope coefficients to vary and be heterogeneous (error

variances are allowed to vary across countries) in the long-run as well as in the short-run.

The third approach is the ‘pooled mean group’ (PMG) estimation introduced by Pesaran

et al (1997; 1999) which combines both pooling and averaging. The PMG estimator, as with the

MG estimator, allows the intercept, short-run coefficients and error variances to differ across

countries, but restricts the long-run (co-integrating) coefficients to be the same across countries

as with the DFE estimation method.

One advantage of the PMG over the traditional DFE is that PMG allows for the short-run

dynamic specification to differ from country to country. In terms of selecting the preferred

estimates between MG and PMG, one could use the Hausman test to see whether there is any

significant difference between the two sets of estimates. The null hypothesis of such a hypothesis

test can be defined as the difference between the MG and PMG estimates being insignificant. If

the null hypothesis is not rejected, we conclude that there is no significant difference between the

two set of estimates and select PMG estimates as they are efficient estimates. If the null

hypothesis is rejected, then we conclude that there is significant difference between the two set

of estimates. One possible solution in this situation is to use the average of the two estimators.

5. Empirical Results

We use models (1) to (3) for our analysis by replacing Y by GDP and the matrix X consisting of

variables CF, DB, TR and P. Initial analysis indicates that an ARDL(1,1,1,1,1) dynamic panel

specification is suitable for our study.

15

Table 5 presents the three sets of alternative pooled estimates, namely, (1) MG estimates

which imposes no restrictions; (2) PMG estimates, which imposes common long-run effects;

and, (3) the DFE estimates which constrains all of the slope coefficients and error variances to be

the same. As can be seen, the signs of the coefficients are mostly the same, however, some

differences can also be noticed between the three sets of estimates; the standard errors of the

estimated long-run coefficients are also generally lower for the PMG estimates than for the MG

and DFE estimates. Across all three sets of estimates, the speed of adjustment (the error

correction term) is

Table 5 Panel ARDL Estimation Results from 40 HIPCs, 1970-20075

VariablePooled Mean Group (PMG) Mean Group (MG)

Dynamic Fixed Effect (DFE)

Error Correction term -0.039 -0.253 -0.046Standard error (0.01) (0.06) (0.02)

p-value (0.00) (0.00) (0.01)

Short-runCapital formation 59.78 39.45 -3.35

Standard error (34.54) (50.55) (35.30)p-value (0.08) (0.44) (0.92)

Debt -218.36 -180.18 -94.29Standard error (25.39) (27.06) (21.38)

p-value (0.00) (0.00) (0.00)Trade -0.96 -0.95 -0.90

Standard error (0.32) (0.30) (0.37)p-value (0.00) (0.00) (0.02)

Population ( x 106) 124.00 -107.70 -21.30Standard error (287.60) (98.40) (17.00)

p-value (0.67) (0.27) (0.21)

Long-runCapital formation 2855.324 2384.174 1882.189

Standard error (659.76) (1003.36) (835.06)p-value (0.00) (0.02) (0.02)

Debt -559.224 -393.977 -315.779Standard error (118.99) (158.13) (152.45)

p-value (0.00) (0.01) (0.04)Trade 4.778 -1.658 7.829

Standard error (2.03) (4.16) (5.63)p-value (0.02) (0.69) (0.17)

Population ( x 106) 30.400 264.900 20.900Standard error (10.20) (147.70) (18.80)

p-value (0.00) (0.07) (0.27)Constant 17.20 -25.79 3.69

Standard error (9.62) (62.15) (11.81)p-value (0.07) (0.68) (0.76)

5 We also included Education as a variable in the model and the results are presented in Table A2 in the Appendix. Since the estimated coefficient for education has the incorrect sign and is insignificant, education was dropped from the model estimation.

16

negative and statistically significant. As expected, from Econometric theory, the MG error

correction estimate presented in Table 5 indicates a much faster adjustment than the PMG or

DEF error correction estimates (-0.253 versus -0.039 and -0.046). The individual country short-

run PMG estimates are given in Table A2 of the Appendix to this paper (available upon request

from the authors). For a long-run relationship to exist, we require i to be non-zero. The results

show that all but one of the individual error correction coefficients is negative and less than one

in absolute value. This means that, for a majority of the individual countries, the hypothesis of no

long-run relationship would not be rejected (for theoretical results, see Pesaran et al, 1999).

Comparing the long-run standard errors of the three sets of estimates, we could see in the

case of PMG that imposing a long-run homogeneity reduces the standard errors of the long-run

coefficients. Furthermore, all of the estimated short-run and long-run debt coefficients are

negative and statistically significant at the 5 percent level.

We use the Hausman test to test the null hypothesis of no difference between the MG and

PMG estimators6. The value of the test statistic is 15.18 with a p-value of 0.002, indicating that

we are unable to accept the null hypothesis that the MG and PMG estimates are the same.

Capital Formation as a proportion of GDP

As expected, across the three estimation methods, in general, the capital formation variable has a

positive impact on GDP in the short run as well as in the long run. This means that higher levels

of capital formation, as a proportion of GDP, would have increased the level of GDP in the

HIPCs. At the 10 percent level of significance, the PMG estimate of capital formation variable is

statistically significant in both the short- and long-run. The long-run MG and DFE estimates for

the capital formation variable are statistically significant at the 5 percent level; but both the short

run MG and DFE estimates are statistically insignificant.

6 We use the STATA software for estimation.17

Debt as a proportion of GDP

Across the three estimation methods, the debt variable has a negative and statistically significant

influence on GDP in the short run as well as in the long run and supports prior expectations. This

means that higher levels of debt, as a proportion of GDP, would have reduced the level of GDP

in the HIPCs. At the 5 percent level of significance, all the PMG, MG and DFE coefficient

estimates are statistically significant in both the short-run and long-run.

This is an interesting result as it is in line with the ‘debt overhang’ hypothesis, which

states that a country experiences debt overhang when its stock of external debt exceeds its ability

to repay its debt. This can have negative impact on economic growth of a debt ridden country as

a large proportion of its output is used to repay debts to foreign lenders, which consequently

creates disincentive to invest (Krugman, 1988; and Sachs, 1989).

This result raises important policy questions. If this relationship is true of the past, and

we expect it to remain true into the future, then a reduction of HIPCs’ debt burden should

increase their level of GDP. This, however, does not say that taking on external debt is bad if

managed well. For example, some of the East Asian economies, especially, the ‘Tiger

Economies’ were able to reduce poverty and enhance economic growth through efficient

utilisation of foreign debts.

Total Trade as a proportion of GDP

The three sets of short-run negative estimates for the trade variable coefficient suggest that

increasing total merchandise exports as a proportion of GDP has a statistically significant

negative effect, in the short-run, on GDP in the HIPCs. The long-run PMG coefficient estimate

for the trade variable is positive and statistically significant but the long-run MG and DFE

coefficient estimates for trade variable are statistically insignificant.

This result could give rise to a change in development assistance policy. It also suggests

that the best way to promote HIPC countries’ economic performance in the long-run is not to

give more aid, but to promote more exports.

18

Population

None of the short-run population coefficients are statistically significant. However, the long-run

PMG and MG population coefficient estimates are both positive and statistically significant. This

means that, in the long-run, an increase in population would impact positively on the GDP,

which may be due to the increase in the work force or human capital.

19

6. Policy Implications

The implications of the above findings, and the review of the literature for economic policy, are

that there is still a significant role for government to play in smoothing out the short term

fluctuations of the business cycles. During recessions, it is advisable for governments to increase

spending, via ‘Fiscal Policy’, even if this means increasing debt levels. In the short term, this

will stimulate economic activity and reduce unemployment further than would occur if the entire

burden were taken on just by ‘Monetary Policy’. In the longer term, after the short term

economic benefits from government spending have been absorbed and the economy has

recovered (even started to heat up), government debt levels can be brought to more sustainable

levels via spending cuts and/or tax increases in order to reduce inflation pressures, and in

preparation for future downturns where increased spending/ debt levels may again be required.

For countries already under significant debt strain (beyond the optimal “threshold”,

which Pattillo et al. (2002) and Clements et al. (2003) put as low as 17-25 percent of GDP,

beyond which additional debt has a negative growth impact), where additional government

spending may worsen an economy’s situation without any corresponding benefit to economic

growth; implying that economic reform and investment must be coupled with debt relief. This

way, economic growth can be stimulated without the need for growth-hindering austerity. This

has significant implications for groups such as the IMF and the World Bank in terms of how they

assist HIPCs in the most effective and efficient manner ergo long-term gains with minimal short-

term pain.

It is also important for groups like the IMF and World Bank to appreciate that developing

and HIPCs often need to accumulate debt in order to supplement their inadequate domestic

funding sources. Consequently, policy advice to these countries must not unduly hinder their

ability to borrow money – the negative growth impacts of this may outweigh the benefits of the

advised reforms, especially in the short term.

20

In terms of the specific types of reforms that can stimulate growth in HIPCs, especially in the

longer term, these can include:

Export- led growth: instead of just providing foreign aid, export-oriented industries should be

supported, including agriculture and labour-intensive manufacturing (industries in which

developing countries often have an advantage), but also in more sophisticated industries such

as capital-intensive manufacturing and technology or service-based industries. Certain

developing countries may not possess the natural advantages to succeed in these latter

industries. Consequently, support for these industries must focus on improving the education

and skills of the local workforce to adequately equip them for success in these industries.

Otherwise, support for these industries may simply divert resources away from industries in

which the country already has an advantage, with no long-term benefit – the country may not

develop an advantage in these new sophisticated industries without a workforce that is

adequately skilled and educated. Argentina is an example of a country that diverted resources

away from its advantages in export-oriented agriculture in favour of import substitution,

without adequately educating and skilling its workforce for the task. Consequently, after

starting the 20th century as one of the world’s wealthiest countries, Argentina suffered

perpetual decline.

Improved macroeconomic policy, including more sustainable monetary policy that generates

more sustainable current account and fiscal balances. This, in turn, will create an

environment less likely to suffer capital flight, and, therefore, less likely to require significant

debt accumulation to support.

Capital formation also has a positive relationship with GDP. Consequently, policy should

encourage investment in capital, especially that which supports traditional export-oriented

industries, but also more sophisticated capital-, technology-, and service-intensive industries.

Investment in capital for service-intensive industries will be most effective when combined

with improved education and skilling of the workforce to suit these industries.

21

Population has a long run positive impact on GDP. Although, given the fact that many

developing or HIPCs already have large young and/ or working-age populations, rather than

ageing populations, population growth may not need to be further encouraged, especially if it

adds to pre-existing unemployment issues. Skilled immigration can be valuable to support

the more sophisticated capital-intensive industries but it is important that this immigration

only temporarily supplements, not permanently replaces, the local workforce until the local

population becomes sufficiently educated and skilled to undertake this work for themselves.

Otherwise there is a risk of significant entrenched unemployment and inequality.

7. Conclusion

22

There is a significant amount of literature that examines the impact of external debt on economic

growth in highly indebted poor countries (HIPCs). This literature was largely consistent in its

findings that debt can have a positive short term economic growth impact up to a certain point.

This is consistent with typical Keynesian recommendations that would result in greater

government spending (and potentially debt accumulation) during recessions, and reduced

spending during booms, as a way of smoothing out the phases of the business cycle in the short

term, thereby reducing the occurrence and subsequent hardships of significant fluctuations in

unemployment and inflation. Furthermore, debt was often necessary for HIPCs because their

own economic growth, incomes, and savings levels were insufficient to finance their

development/ national goals; and their generally commodity-based exports were not sufficient to

finance their more expensive capital intensive imports. In the long term however, and beyond a

certain “threshold” debt level, debt appears to have negative economic growth and current

account effects, driven by the lower efficiency of investment as debt increases, as well as capital

and labour productivity effects and “crowding out” effects. Consequently, for HIPCs, the best

way to ensure appropriate lending or investment, and economic reforms (for example, income-

generating activities, export-oriented trade policy, appropriate macroeconomic policy) to actually

have positive economic growth and social impacts is to combine the loans and reforms with debt

relief so that new investment is not impeded by debt and illiquidity. Furthermore, government

spending (and potentially debt accumulation) can play a valuable role during booms and

recessions, even if it does not increase (or even decreases) long term growth, as long as debt

levels are reduced to (or below) the “threshold” level over the long term.

23

We examined short-run and long-run relationships between external debt and economic

growth in 40 HIPCs over the period 1970-2007 with the aid of the growth accounting process. In

addition, the impacts of capital formation, trade and population growth on economic growth in

these countries was also examined. We used panel data estimation of an ARDL model. Our

analysis reveals that GDP and merchandise export growth accelerated fastest since 2000, while

debt levels have fallen since 2000. Specifically, the growth rate of debt in these countries has

been falling since 1970, and was actually largely negative since 2000. The estimation results

indicate firstly that, capital formation’s share of GDP has a positive impact on HIPC’s GDPs in

the short-run as well as in the long-run. Secondly, debt as a share of GDP has a negative

influence in the short-run as well as in the long-run. This is consistent with the debt overhang

hypothesis that a country experiences debt overhang when its stock of external debt exceeds its

ability to repay its debt, which causes a large proportion of their GDP to have to be committed to

re-paying foreign lenders, which disincentives investment. Therefore, reducing HIPC’s debt

would be good for their GDP, but debt can still be good for GDP up to a certain point, for

example, East Asia’s efficient use of foreign debt to reduce poverty and enhance economic

growth. Thirdly, in the long-run, the merchandise trade as a percentage of GDP has a positive

influence on GDP. This has implications for development assistance policy in that it should

encourage exports, rather than just delivering foreign aid. Fourthly and finally, population

increase has a positive influence on the economic growth in the HIPCs, possibly due to its

workforce/ human capital impacts.

24

References

Abbott, G. (1993). Debt Relief and Sustainable Development in Sub-Saharan Africa. Edward Elgar Publishing, USA.

Abrego, L., and Ross, D. (2001). “Debt Relief Under the HIPC Initiative: Context and Outlook for Debt Sustainability and Resource Flow”, IMF Working Paper Series, Working Paper No. 144, IMF, Washington.

Adams, F.G., E.P. Sanchez and M.E. Adams (1983) “Can Latin America carry its International Debt? A prospective analysis using the Wharton Latin American Debt Simulation Model”, Journal of Policy Modelling 5(3): 419-441.

Adegbite, E.O., Ayadi, F.S., and Ayadi, O.F. (2008). “The Impact of Nigeria’s External Debt on Economic Development”, International Journal of Emerging Market 3(3): 285-301.

Altvater, E. (1991). The Poverty of Nations: A Guide to the Debt Crisis from Argentina to Zaire, Zed Books, London.

Baltagi, B.H., Griffin, J.M, and Xiong, W. (2000). “To Pool or Not To Pool: Homogeneous Versus Heterogeneous Estimators Applied to Cigarette Demand”, Review of Economics and Statistics 82(1): 117-126.

Barro, R. (1989). “Economic Growth in a Cross-Section of Countries”, NBER Working Paper Series, No. 3120, National Bureau of Economic Research, USA, September 1989.

Barro, R., and Lee, J. (1994). “Sources of Economic Growth”, Carnegie-Rochester Conference Series on Public Policy 40(9): 1-36.

Berensmann, K. (2004). “New Ways of Achieving Debt Sustainability beyond the Enhanced HIPC Initiative”, Intereconomics 39(6): 321-330.

Birdsall, N., Williamson, J., and Deese, B. (2002). Delivering on Debt Relief: From IMF Gold to a New Aid Architecture. Co-published by the Center for Global Development and the Institute for International Economics, Washington.

Claessens, S. (1990). “The Debt Laffer Curve: Some Estimates”, World Development 18(12):

Clements, B., Bhattacharya, R., and Nguyen, T. (2003). “External Debt, Public Investment, and Growth in Low-Income Countries”, IMF Working Paper No. 249, WP/03/249, IMF, Washington.

De Macedo, J.B. (1990). ‘External liberalisation with ambiguous public response: the experience of Portugal’. Cambridge University Press.

Dickey, D.A., and Fuller, W.A. (1979). "Distribution of the Estimators for Autoregressive Time Series with a Unit Root", Journal of the American Statistical Association 74 (366): 427–431.

Dickey, D.A., and Fuller, W.A. (1981). “Distribution of the Estimators for Autoregressive Time Series with a Unit Root”, Econometrica 49: 1057-1072.

Easterly, W. (2001). “The Lost Decades: Developing Countries’ Stagnation in Spite of Policy Reform 1980-1998”, Journal of Economic Growth 6: 135-157

Faini, R. and DeMelo, J. (1990). “Adjustment, Investment, and Real Exchange Rate in Developing Countries”, Economic Policy 11: 492-512.

Filatov, V.S., and R.P. Mattione (1985) “Latin America's Recovery from Debt Problems: An Assessment of Model-based Projections”, Journal of Policy Modelling 7(3): 491-524.

Fonchamnyo, D.C. (2009). “Debt Relief Incentives in Highly Indebted Poor Countries (HIPC): An Empirical Assessment”, International Advances in Economic Research 15: 322-335.

25

Fosu, A.K. (1999). “The External Debt Burden and Economic Growth in the 1980s: Evidence from Sub-Saharan Africa”, Canadian Journal of Development Studies 20(2): 307-318.

Froot, K. (1989). “Buybacks, Exit Bonds, and the Optimality of Debt and Liquidity Relief”, International Economic Review 30: 49-70.

Fry, M.J. (1989). “Foreign Debt Instability: An Analysis of National Saving and Domestic Investment Response to Foreign Debt Accumulation in 28 Developing Countries”, Journal of International Money and Finance 8: 315-344.

Giersch, H. (1985). International Debt Problem: Lesson for the Future, Institute of World Economics, University of Kiel, Westview Press.

Gómez-Puig, M., and S. Sosvilla-Rivero (2015) “The causal relationship between debt and growth in EMU countries”, Journal of Policy Modelling 37(6): 974-989.

Hadri, K. (2000). “Testing for Stationarity in Heterogeneous Panel Data”, The Econometrics Journal 3(2): 148–161.

Hallett, A.H., and J.C.M. Olivia (2015) “The Importance of Trade and Capital Imbalances in the European Debt Crisis”, Journal of Policy Modelling 37(2): 229-252.

Hameed, A., Ashraf, H., and Chaudhary, M.A. (2008). “External Debt and Its Impact on Economic and Business Growth in Pakistan”, International Research Journal of Finance and Economics 20: 132-140.

Hoffman, B. and H. Reisen (1991). “Some Evidence on Debt-Related Determinants on Investment and Consumption in Heavily Indebted Countries”, Weltwirschaftliches Archive 127(2): 280-297.

Im, K.S., Pesaran, M.H., and Shin, Y. (2003): “Testing for Unit Roots in Heterogeneous Panels”, Journal of Econometrics 115: 53-74.

Jayaraman, T.K., and Lau, E. (2009). “Does External Debt Lead to Economic Growth in Pacific Island Countries”, Journal of Policy Modelling 31: 272-288.

IMF (2014). International Monetary Fund, “Debt Relief under the Heavily Indebted Poor Countries (HIPC) Initiative”, Factsheet, September 30, 2014, available at: ttps://www.imf.org/external/np/exr/facts/hipc.htm-accessed on 24 March, 2015.

Kao, C. and McCoskey, S. (1998) “A Residual-Based Test of the Null of Cointegration in Panel Data”, Econometric Reviews 17: 57-84.

Kraay, A. and Nehru, V., (2006). “When Is External Debt Sustainable?”, The World Bank Economic Review 20(3): 341-365.

Krueger, A.O. (1987). “Origins of the Developing Countries’ Debt Crisis 1970 to 1982”, Journal of Development Economics 27: 165-187.

Krugman, P. (1988). “Financing vs. Forgiving a Debt Overhang: Some Analytical Notes”, Journal of Development Economics 29: 253-268.

Kwack, S.Y. (1983) “Developments in and prospects for external debt position and burden of developing countries: The case of Korea”, Journal of Policy Modelling 5(3): 443-459.

Leipziger, D. and Allen, M., (2008) “Heavily Indebted Poor Countries (HIPC) Initiative and Multilateral Debt Relief Initiative (MDRI) – Status of Implementation”, International Development Association and International Monetary Fund.

Levin, A., Lin, C.F., and Chu, C. (2002). “Unit Root Tests in Panel Data: Asymptotic and Finite Sample Properties”, Journal of Econometrics 108: 1-24.

Maddala, G.S., and Wu, S. (1999). “A comparative study of unit root tests with panel data and New Simple Test”, Oxford Bulletin of Economics and Statistics 61: 631-652

26

Maghyereh, A., and Hashemite, U., (2003) “External Debt and Economic Growth in Jordan: The Threshold Effect”, Economia Internazionale/International Economics 56(3): 337-355.

Mitze, T., and F. Matz (2015) “Public Debt and Growth in German Federal States: What can Europe Learn?”, Journal of Policy Modelling 37(2): 208-228.

Pattillo, C., Poirson, H., and Ricci, L. (2002). “External Debt and Growth”, IMF Working Paper No. 2, IMF, Washington.

Pesaran, M.H., and Shin, Y., (1999). “An autoregressive distributed lag modelling approach to cointegration analysis”, Chapter 11 in Econometrics and Economic Theory in the 20th Century: The Ragnar Frisch Centennial Symposium, Strom, S. (ed.), Cambridge University Press, Cambridge.

Pesaran, M.H., and Shin, Y. (2003). “Testing for Unit Roots in Heterogeneous Panels”, Journal of Econometrics 115: 53-74.

Pesaran, M.H., Shin, Y., and Smith, R.P. (1997). “Estimating Long-run Relationships in Dynamic Heterogeneous Panels”, DAE Working Papers Amalgamated Series 9721.

Pesaran, M.H., Shin, Y. and Smith, R.P. (1999). “Pooled Mean Group Estimation of Dynamic Heterogeneous Panels”, Journal of the American Statistical Association 94: 621-634.

Pesaran, M.H., Shin, Y. and Smith, R.P. (2001). “Bounds Testing Approaches to the Analysis of Level Relationships”, Journal of Applied Econometrics 16: 289–326.

Pesaran, M.H., and Smith, R.P. (1995) “Estimating Long-run Relationships from Dynamic Heterogeneous Panels”, Journal of Econometrics 68: 79-113.

Phillips, P.C.B., and Perron, P. (1988). “Testing for a Unit Root in Time Series Regression”, Biometrika 75(2): 335–346.

Sachs, J. (1989). “The Debt Overhang of Developing Countries”, in Calvo, G, R Findlay, P Kouri and J Macedo, eds., Debt, Stabilisation and Development: Essays in Memory of Carlos Diaz Alejandro, Oxford, Basic Blackwell, pp. 80-102.

Sachs, J.D., Botchwey, K. & Cuchra, M. (1999). ‘Implementing Debt Relief for the HIPCs’. Columbia University Academic Commons.

Siddique, M. A. B. (1996). “The External Debt Problem of Sub-Saharan Africa: 1971-1990”, the South African Journal of Economics 64(2): 100-124.

27

APPENDIX

Individual country unit root test and test of cointegration

First we apply the Augmented Dicky Fuller (ADF) test (Dickey and Fuller, 1979; 1981) to test

the stationarity of each of the five variables GDP, CF, DB, TR and P for each country. The null

hypothesis of the ADF test is that the time series has a unit root (or the time series is non-

stationary). Table A1 presents the unit root test results. As can be seen, the p-value for testing the

null hypothesis of the existence of a unit root is mostly less than the level of significance, α =

0.05, thus we reject the null hypothesis. This means that, almost all time series are stationary,

that is I(0), except GDP for Kyrgrz Republic and Tanzania; capital formation for Liberia; debt

for Eritrea, Madagascar and São Tomé and Príncipe; and, population for Benin and Comoros.

The individual country short-run PMG estimates and their standard errors for equation (3) are

given in Table A2.

Table A3 presents the corresponding estimation results in Table 5 with Education (education

expenditure as a proportion of GDP) included as an additional variable in equation (3). Across

the three sets of short-run estimates, the coefficient attached to the education variable is mostly

negative but statistically insignificant. The long run PMG coefficient estimate for education is

negative and statistically significant, and the long-run MG and DFE coefficient estimates of

education are insignificant. These results suggest that expenditure on education as a proportion

of GDP is detrimental to the level of GDP in the HIPCs. The direction of this relationship is

markedly different from rational expectation.

28

Table A1 Value of the ADF test statistic and its p-value (in parentheses) from the Unit root test for each variable in each country

Country1 Benin -5.841 (0.000) -7.910 (0.000) -7.954 (0.000) -6.782 (0.000) -1.832 (0.355)2 Bolivia -4.860 (0.000) -6.080 (0.000) -4.560 (0.000) -5.330 (0.000) -6.266 (0.000)3 Burkina Faso -6.830 (0.000) -7.640 (0.000) -4.890 (0.000) -7.030 (0.000) -5.688 (0.000)4 Burundi -6.950 (0.000) -8.930 (0.000) -4.500 (0.000) -5.870 (0.000) -6.255 (0.000)5 Cameroon -4.350 (0.000) -6.840 (0.000) -5.340 (0.000) -6.080 (0.000) -5.718 (0.000)6 Central African Republic -6.874 (0.000) -7.559 (0.000) -7.190 (0.000) -6.874 (0.000) -6.264 (0.000)7 Chad -5.116 (0.000) -5.004 (0.000) -4.949 (0.000) -9.090 (0.000) -3.768 (0.007)8 Comoros -4.694 (0.000) -6.827 (0.000) -4.553 (0.000) -2.715 (0.009) 0.767 (0.991)9 Congo, Dem. Rep. -6.599 (0.000) -9.103 (0.000) -5.777 (0.000) -8.223 (0.000) -6.621 (0.000)10 Congo, Rep. -5.341 (0.000) -5.313 (0.000) -5.222 (0.000) -6.472 (0.000) -6.341 (0.000)11 Cote d'Ivoire -6.240 (0.000) -5.513 (0.000) -2.953 (0.004) -4.896 (0.000) -8.735 (0.000)12 Eritrea -2.145 (0.035) -6.018 (0.000) -1.840 (0.065) -4.646 (0.000) -6.250 (0.000)13 Ethiopia -6.600 (0.000) -8.385 (0.000) -4.315 (0.000) -2.990 (0.004) -4.098 (0.000)14 Gambia, The -6.609 (0.000) -5.751 (0.000) -5.367 (0.000) -8.218 (0.000) -6.319 (0.000)15 Ghana -5.653 (0.000) -6.260 (0.000) -4.733 (0.000) -4.828 (0.000) -6.603 (0.000)16 Guinea -2.873 (0.006) -3.936 (0.001) -4.067 (0.000) -3.611 (0.001) -6.231 (0.000)17 Guinea-Bissau -8.945 (0.000) -6.977 (0.000) -5.602 (0.000) -6.946 (0.000) -6.272 (0.000)18 Guyana -4.145 (0.000) -6.102 (0.000) -5.061 (0.000) -5.586 (0.000) -6.530 (0.000)19 Haiti -7.104 (0.000) -5.320 (0.000) -5.469 (0.000) -9.381 (0.000) -6.480 (0.000)20 Honduras -4.888 (0.000) -5.599 (0.000) -4.420 (0.000) -6.240 (0.000) -6.161 (0.000)21 Kyrgyz Republic -0.509 (0.480) -3.112 (0.004) -2.305 (0.025) -3.332 (0.003) -6.288 (0.000)22 Liberia -7.494 (0.000) -0.855 (0.297) -3.180 (0.002) -6.478 (0.000) -6.037 (0.000)23 Madagascar -6.189 (0.000) -7.141 (0.000) -1.812 (0.067) -6.108 (0.000) -6.655 (0.000)24 Malawi -7.845 (0.000) -8.419 (0.000) -5.463 (0.000) -7.090 (0.000) -6.208 (0.000)25 Mali -5.392 (0.000) -9.632 (0.000) -7.156 (0.000) -7.956 (0.000) -6.278 (0.000)26 Mauritania -6.127 (0.000) -9.789 (0.000) -4.576 (0.000) -8.257 (0.000) -6.337 (0.000)27 Mozambique -4.889 (0.000) -7.221 (0.000) -4.088 (0.000) -3.167 (0.003) -6.391 (0.000)28 Nepal -6.414 (0.000) -7.499 (0.000) -3.045 (0.003) -5.870 (0.000) -6.010 (0.000)29 Nicaragua -5.893 (0.000) -7.565 (0.000) -5.241 (0.000) -6.015 (0.000) -6.253 (0.000)30 Niger -6.435 (0.000) -5.957 (0.000) -5.201 (0.000) -6.620 (0.000) -4.797 (0.000)31 Rwanda -7.253 (0.000) -9.864 (0.000) -6.996 (0.000) -9.781 (0.000) -6.204 (0.000)32 Sao Tome and Principe -4.317 (0.000) a a -1.630 (0.096) a a -6.345 (0.000)33 Senegal -6.517 (0.000) -7.163 (0.000) -8.768 (0.000) -7.537 (0.000) -6.570 (0.000)34 Sierra Leone -7.088 (0.000) -8.371 (0.000) -4.767 (0.000) -7.777 (0.000) -6.299 (0.000)35 Somalia -5.319 (0.000) -10.053 (0.000) -4.245 (0.000) -4.658 (0.000) -5.009 (0.000)36 Sudan -3.574 (0.001) -5.086 (0.000) -5.692 (0.000) -5.840 (0.000) -6.805 (0.000)37 Tanzania -1.657 (0.091) -2.797 (0.009) -3.167 (0.003) -3.119 (0.004) -6.120 (0.000)38 Togo -6.088 (0.000) -5.621 (0.000) -5.382 (0.000) -6.565 (0.000) -6.300 (0.000)39 Uganda -5.553 (0.000) -7.667 (0.000) -4.827 (0.000) -5.973 (0.000) -5.947 (0.000)40 Zambia -4.020 (0.000) -7.190 (0.000) -5.224 (0.000) -6.237 (0.000) -6.349 (0.000)

GDP Capital formation Debt PopulationTrade

a: Insufficient observations; b: near-singular; highlighted cells represent insignificant coefficient estimates at the 5 percent level.

29

Table A2 ARDL Estimation Results, Individual HIPCs, 1970-2007Country1 -0.039 (0.171) 79.48 (0.520) -293.42 (0.000) -0.130 (0.876) -38.70 (0.660) 10.25 (0.604)2 -0.080 (0.055) -198.32 (0.170) -307.58 (0.000) 0.550 (0.750) 718.30 (0.203) -65.28 (0.393)3 -0.029 (0.410) -170.79 (0.216) -504.67 (0.000) 1.479 (0.250) -99.70 (0.001) 19.34 (0.405)4 -0.014 (0.041) 55.95 (0.361) -55.19 (0.002) 0.324 (0.478) -17.70 (0.426) 4.61 (0.228)5 -0.040 (0.002) -373.49 (0.203) -487.70 (0.000) -7.670 (0.000) -105.40 (0.002) 47.35 (0.021)6 -0.031 (0.055) -97.13 (0.540) -314.44 (0.000) -0.528 (0.428) -135.30 (0.396) 18.56 (0.132)7 -0.034 (0.006) -189.43 (0.000) -308.20 (0.000) -0.149 (0.556) 24.50 (0.459) -3.75 (0.613)8 -0.083 (0.002) -17.19 (0.878) -426.18 (0.000) -0.526 (0.646) 10702.00 (0.104) -88.32 (0.214)9 -0.038 (0.109) -29.56 (0.811) -70.30 (0.013) -1.074 (0.096) -24.10 (0.184) -3.19 (0.878)10 -0.009 (0.650) 199.43 (0.363) -356.23 (0.000) 1.343 (0.203) -1635.50 (0.000) 160.25 (0.000)11 -0.092 (0.011) 703.94 (0.185) -314.49 (0.000) -4.140 (0.051) 37.80 (0.810) 29.95 (0.612)12 0.006 (0.639) 41.12 (0.301) -157.08 (0.000) -0.090 (0.808) 23.50 (0.737) 16.79 (0.328)13 -0.011 (0.074) -23.13 (0.786) -43.26 (0.008) -4.326 (0.000) 2.91 (0.388) -17.29 (0.088)14 -0.048 (0.007) 39.80 (0.718) -150.24 (0.000) 0.118 (0.713) -121.70 (0.429) 3.68 (0.747)15 -0.018 (0.220) 162.45 (0.289) -258.69 (0.000) -1.031 (0.150) -63.20 (0.200) 27.13 (0.127)16 -0.071 (0.195) 231.80 (0.324) -433.11 (0.000) -1.457 (0.104) -77.80 (0.126) 13.75 (0.312)17 -0.004 (0.336) 36.54 (0.367) -43.40 (0.000) -0.660 (0.009) -249.10 (0.331) 14.57 (0.044)18 -0.020 (0.448) 14.52 (0.926) -106.77 (0.000) 0.055 (0.893) -1456.60 (0.783) 32.54 (0.091)19 -0.083 (0.014) -36.36 (0.851) -648.58 (0.000) -0.288 (0.698) -13.20 (0.728) -17.21 (0.264)20 -0.034 (0.138) 378.14 (0.085) -412.86 (0.000) -1.442 (0.197) -620.80 (0.000) 112.84 (0.000)21 -0.091 (0.023) 3.46 (0.979) -291.17 (0.000) 1.881 (0.004) 645.80 (0.020) -31.73 (0.122)22 -0.009 (0.085) 210.12 (0.554) -26.86 (0.002) -0.173 (0.655) -48.40 (0.511) 17.83 (0.008)23 -0.044 (0.030) 343.72 (0.147) -50.70 (0.157) -1.059 (0.157) -94.60 (0.077) 26.71 (0.125)24 -0.013 (0.144) -32.05 (0.418) -74.49 (0.000) -0.694 (0.007) -21.70 (0.370) 3.89 (0.669)25 -0.050 (0.095) -24.69 (0.813) -129.80 (0.001) -1.061 (0.055) -23.40 (0.041) -2.82 (0.810)26 -0.035 (0.001) -44.59 (0.109) -216.18 (0.000) -0.323 (0.265) -897.30 (0.000) 69.94 (0.000)27 -0.034 (0.030) -112.94 (0.316) -10.25 (0.605) -5.214 (0.000) 7.89 (0.746) 10.63 (0.350)28 -0.041 (0.042) -156.18 (0.078) -213.16 (0.000) 0.203 (0.716) -75.30 (0.029) 2.36 (0.822)29 -0.020 (0.038) 643.61 (0.000) -48.81 (0.000) -5.437 (0.000) -145.50 (0.467) 53.31 (0.017)30 -0.046 (0.045) 222.48 (0.057) -158.61 (0.002) -0.139 (0.862) -42.80 (0.404) 1.67 (0.926)31 -0.071 (0.002) -1.17 (0.994) -106.36 (0.000) -1.739 (0.001) -13.20 (0.301) -13.90 (0.155)32 a33 -0.053 (0.035) 84.80 (0.709) -439.30 (0.000) -2.602 (0.008) -161.90 (0.267) 40.87 (0.227)34 -0.006 (0.729) -100.90 (0.535) -104.58 (0.000) 0.342 (0.545) -49.70 (0.506) 9.69 (0.283)35 -0.020 (0.080) -67.18 (0.170) -79.38 (0.000) -0.513 (0.002) -35.80 (0.092) 10.75 (0.036)36 -0.111 (0.004) 321.06 (0.172) -173.96 (0.000) -0.413 (0.846) 104.40 (0.083) -88.28 (0.021)37 -0.027 (0.059) -116.50 (0.311) -77.71 (0.002) -0.518 (0.262) -6.34 (0.704) -16.70 (0.425)38 -0.065 (0.002) 68.20 (0.364) -186.84 (0.000) 0.055 (0.895) 96.60 (0.487) -23.27 (0.302)39 -0.005 (0.863) 260.49 (0.252) -273.89 (0.000) -1.540 (0.133) -26.20 (0.494) 18.48 (0.216)40 0.040 (0.078) 21.97 (0.888) -161.53 (0.000) 1.091 (0.295) -1226.70 (0.000) 264.85 (0.000)

Mean -0.038 59.78 -218.36 -0.961 124.00 17.201

GDP Capital formation TradeDebt Population Constant

a: Insufficient observations.

30

Table A3 Panel ARDL Estimation Results (including Education) from 40 HIPCs, 1970-2007

VariablePooled Mean Group (PMG) Mean Group (MG)

Dynamic Fixed Effect (DFE)

Short-runError Correction -0.07 -0.34 -0.04

(0.02) (0.15) (0.02)(0.00) (0.03) (0.01)

Capital formation 81.88 13.94 16.82(39.87) (53.73) (41.54)

(0.04) (0.80) (0.69)Debt -233.18 -167.60 -98.63

(27.02) (39.96) (24.36)(0.00) (0.00) (0.00)

Trade -0.83 -0.94 -1.03(0.33) (0.33) (0.46)(0.01) (0.01) (0.03)

Education -5.23 -10.61 -10.29(4.53) (7.17) (7.60)(0.25) (0.14) (0.18)

Population ( x 106) 169.00 -368.10 -18.10(428.90) (261.90) (15.20)

(0.69) (0.16) (0.23)Constant 49.64 -49.36 11.92

(9.99) (141.02) (14.53)(0.00) (0.73) (0.41)

Long-runCapital formation 1016.842 1420.102 1945.135

(146.68) (442.78) (1003.79)(0.00) (0.00) (0.05)

Debt -95.197 -316.774 -460.706(13.06) (168.56) (271.40)

(0.00) (0.06) (0.09)Trade 2.183 -15.315 12.923

(0.64) (11.73) (8.60)(0.00) (0.19) (0.13)

Education -56.393 36.123 -105.000(13.01) (89.97) (95.34)

(0.00) (0.69) (0.27)

Population ( x 106) -4.890 -31.200 19.000(2.68) (313.70) (20.50)(0.07) (0.92) (0.35)

For each estimate the first row in parentheses are the standard errors and the second row provides the corresponding p-values for testing the statistical significance of the estimates.

31