Public debt, institutional quality and growth in sub-Saharan Africa: a threshold analysis

ABSTRACT

Using data for 44 sub-Saharan African countries, this study analyzes the effect of public debt on economic growth and assesses whether institutional quality matters for this effect. The results indicate that, while an increase in public debt has a negative impact on economic growth, this effect is dampened when there is an increase in the quality of institutions as captured by an anti-corruption perception indicator or a government effectiveness indicator. Further, the findings indicate that there is a level of institutional quality above which the effect of an increase in public debt on growth is positive. Ergo, a decline in corruption as well as the perception effect that it engenders, and an improvement in the quality of policymaking should eliminate some of the inefficiencies characteristic of governments in the sub-region and facilitate a positive impact of debt on growth.

KEYWORDS:

• Public debt
• institutions
• growth

JEL:

• F34
• F43
• O11
• O43

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. See Eberhardt and Presbitero (2015) for additional references.

2. See Panizza and Presbitero (2014) for a list of references.

3. See Acemoglu, Johnson, and Robinson (2001, 2002) for a review of the literature.

4. Kim, Yoonhee, and Kim (2017) use data for countries covering the spectrum of economic development.

5. See Acemoglu, Johnson, and Robinson (2004) for a comprehensive description of this framework. We believe that de jure political powers (defined in below) are more important in sub-Saharan Africa than de facto (i.e. institutions-based) political powers. This, and the inclusion of ‘ new resources’ in our modified framework, constitutes the main difference with Acemoglu, Johnson, and Robinson (2004)’s framework.

6. Acemoglu, Johnson, and Robinson (2004) do not provide a clear definition of economic institutions, but only refer to some characteristics of such institutions (e.g. structure of property rights, presence and perfection of markets, etc.). In our analysis, we choose corruption and government effectiveness indicators as proxies of the quality economic institution as these indicators influence the efficiency with which economic resources are allocated.The de facto political powers are defined as political powers that do not necessarily originate in political institutions (e.g. constitution, laws, etc.) but on the use of force, revolt, protest and other unlawful means.

7. The selection of the sample period is dictated by the availability of the data.

8. List of Countries: Angola, Benin, Botswana, Burkina Faso, Burundi, Cameroon, Cape Verde, Central African Republic, Chad, Comoros, Republic of Congo, Democratic Republic of Congo, Djibouti, Equatorial Guinea, Ethiopia, Gabon, The Gambia, Ghana, Guinea, Guinea-Bissau, Cote d’Ivoire, Kenya, Lesotho, Liberia, Madagascar, Malawi, Mali, Mauritania, Mauritius, Mozambique, Namibia, Niger, Rwanda, Sao Tome and Principe, Senegal, Seychelles, South Africa, Sudan, Swaziland, Tanzania, Togo, Uganda, Zambia, Zimbabwe.

9. Some countries display very big change per capita GDP (for instance −50% or 92%). While these growth rates are not unrealistic, given events like civil wars that often lead to dramatic changes in output in SSA, they are atypical and could seriously flaw estimation results. Therefore, the estimations are based on a truncated sample that excludes values above the 99th percentile and below the 1st percentile.

10. This is one of the two techniques used to limit the number of instruments generated in the difference and system GMM estimators as highlighted in Roodman (2009). Thus, we allow a maximum of 2 period lags for instruments.

11. We abstract from a detailed description of the estimator since it has been widely used in several studies (See Arellano and Bover (1995) and Blundell and Bond (1998) for details on the GMM system estimator). Nonetheless, it noteworthy that some issues remain with this estimation technique. For instance, a test for overidentifying restriction might have low power, while weak instruments can arise when the variances of the individual heterogeneity and idiosyncratic errors are the same (see Bun and Windmeijer (2010) for details).

12. We drop outliers for GDP per capita growth and debt-to-GDP ratio, using observations that fall within the 1st and 99th percentiles.

13. The approach involves taking the derivative of the growth equation with respect to debt and setting it equal to zero. The estimate based on the specification in column 1 of Table 2, for example, would be obtained from the resultant equation: $\frac{\mathrm{\Delta }GDPGR}{\mathrm{\Delta }DEBT}=-0.02+0.012\ast corruption=0.$

14. The estimated threshold values are 1.6, 2, and 2.2, based on coefficient estimates on debt and the interaction term for debt and the corruption index in columns (1), (2), (3) respectively, which are the specifications that use the corruption indicator as constructed.

15. Rising public debt also crowds out private investment, although this may be to a limited extent in sub-Saharan Africa as most debt is contracted from external sources and multilateral creditors.

16. A decline in corruption would potentially lead to lesser amounts of disbursed loans being diverted, and more funds being allocated to productive capacity building.

17. The coefficient estimate for government expenditure is negative but not statistically significant in columns (1), (4) and (5) of Table 2. Nevertheless, such an effect, where it exists could be explained by the negative effects of government budget deficits on growth as shown in Woo and Kumar (2015).

18. It is noteworthy that the estimated effect of the interaction between debt and institutions on growth in specifications (1) and (2) is potentially via total factor productivity, and both physical and human capital accumulation.

19. As in Table 2, all explanatory variables, except the lagged dependent variable, are lagged 4 periods across all specifications.

20. The nonlinear combination is estimated at the mean value of the government effectiveness indicator.

21. For the estimate based on the specification in column 2 of Table 4, for example, the resultant equation used is: $\frac{\mathrm{\Delta }GDPGR}{\mathrm{\Delta }DEBT}=-0.022+0.016\ast Govt.$ $effectiveness=0.$

22. We use an adhoc approach by visualizing the data points and dropping the very extreme values of the debt-to-GDP ratio and GDP per capita growth. Thus, we drop observations for debt-to-GDP ratio greater than 300, and observation for GDP per capita growth less than −20 and greater than 20.

23. As noted in section 3.2, the system GMM estimator can be weakened by proliferation of instruments, and although we address this by restricting the lags for instruments to limit the total instruments count, we have some specifications where there are more instruments than number of countries. This occurs when `education’ replaces `population growth’ as a covariate for a robustness check, which drastically reduces the number of countries due to missing observations (Table 2 Column 4; Table 3 Columns 5 and 6; Table 4 Column 5; Table 5 Columns 3 and 6). While the test of joint validity of instruments are considered weaker in these instances, the coefficient estimates for the main variables only serve to confirm the estimates from the other specifications and robustness checks, and do not adversely impact the validity of the main results.