Is public debt harmful towards economic growth? New evidence from South Africa

Figures & data

Figure 1. Debt as a % of GDP and GDP in South Africa (1994–2017).

Table 1. Summary statistics of the time series

	debt_g	debt_n	gdp	inf	inv	tot
Mean	37.20	32.74	2.76	5.77	19.26	0.67
Median	35. 10	32.60	2. 70	5. 90	19.40	0.40
Maximum	51.60	45.70	7. 40	14.10	25.30	7.50
Minimum	26.00	21.60	−6.10	−1.60	15.00	−4.80
Std Dev	7.56	7.06	2. 63	3.22	2.16	2.88
Skewness	0.38	0.25	−0.66	0.52	0.08	0.28
Kurtosis	2.09	2.11	3.76	3.81	3.29	2.70
Jarque Bera	3.60	2.52	5.85	4.39	0.27	1.02
Probability	0.16	0.28	0.05	0.11	0.87	0.59

Table 2. Correlation matrix of the time series

	debt_g	gdp	inf	inv	tot
Debt_g	1
Gdp	−0. 27	1
Inf	−0. 13	−0. 15	1
Inv	−0. 12	−0. 45	0. 26	1
Tot	−0. 09	−0. 03	−0. 15	−0. 05	1
	debt_n	gdp	inf	inv	tot
Debt_n	1
Gdp	−0. 22	1
Inf	−0. 16	−0. 15	1
Inv	−0. 26	−0. 45	0. 26	1
Tot	−0. 07	−0. 15	−0. 15	−0. 061	1

Table 3. Unit root tests results

	Intercept			Intercept and trend
	ADF	PP	DF-GLS	ADF	PP	DF-GLS
Gdp	−2. 09	−2. 13	−2. 19**	−2. 71	−2. 57	−2. 94*
Δgdp	−4. 52***	−5. 99***	−4. 55***	−4. 46**	−7. 82***	−4. 77***
Debt_g	−0. 87	−0. 48	−1. 02	−0. 99	−1. 04	−1. 88
ΔDebt_g	−2. 14	−1. 95	−2. 22**	−1. 95	−3. 49*	−2. 46*
Debt_n	−1. 62	−0. 99	−1. 59	−1. 18	−1. 09	−2. 11
ΔDebt_n	−1. 66	−1. 62	−1. 72*	−1. 67	−3. 21*	−2. 32
Inf	−0.39	−0.79	−1. 10	−3. 38*	−2. 32	−3. 79***
Δinf	−3.64**	−2. 47	−3. 77***	−4. 40**	−5. 14***	−4. 99***
Inv	−2. 07	−1. 49	−1. 93*	−2. 00	−1. 21	−2. 33
Δinv	−4. 66***	−5. 64***	−4. 96***	−2. 60	−2. 44	−2. 82
Tot	−1. 20	−1. 24	−0. 81	−1. 44	−1. 44	−1. 69
Δtot	−3. 51**	−3. 50**	−3. 64***	−3. 55*	−3. 72**	−3. 83***
Critical levels
1% level	−3. 67	−3. 64	−2. 63	−4. 30	−4. 26	−3. 77
5% level	−2. 96	−2. 95	−1. 95	−3. 57	−3. 55	−3. 19
10% level	−2. 62	−2. 61	−1. 61	−3. 22	−3. 20	−2. 89

Table 4. Bounds test for cointegration

Specification	Selected model	F-Statistic	1%	5%	10%
			I(0) I(1)	I(0) I(1)	I(0) I(1)
F(gdp Debt_n)	ARDL (1, 0)	4.34	4.94 5.58	3.62 4.16	3.02 3.51
F(gdp Debt_n, inf, inv, tot)	ARDL(1, 0, 0, 0)	5. 86	3.06 4.15	2.39 3. 38	2.08 3.30
F(gdp Debt_g)	ARDL (1, 0)	5. 77	4.94 5.58	3.62 4.16	3.02 3.51
F(gdp Debt_g, inf, inv, tot)	ARDL (1, 0, 0, 0, 0)	5. 65	3.06 4.15	2.39 3. 38	2.08 3.30

Table 5. Long run and short run ARDL estimates

	F(GDP Debt_n)		F(GDP Debt_n, INF, INV, TOT)		F(GDP Debt_g)		F(GDP Debt_g, INF, INV, TOT)
	Coefficient	p-value	Coefficient	p-value	Coefficient	p-value	Coefficient	p-value
Panel A: Long-run estimates
Debtg	–	–	–	–	−0. 07	0. 03**	−0. 11	0.06*
Debtn	−0.10	0.08*	−0. 13	0. 00***	–	–	–	–
Inf	–	–	−0. 37	0. 00***	–	–	−0. 38	0. 02**
Inv	–	–	−0. 51	0. 00***	–	–	−0. 45	0. 04*
Tot	–	–	−0. 01	0. 00***	–	–	−0. 01	0. 01**
Panel B: Short-run estimates
∆ (Debtg)	–	–	–	–	−0. 10	0. 00***	−0. 20	0.02**
∆ (Debtn)	−0.10	0. 00***	−0. 12	0. 00***	–	–	–	–
∆ (Inf)	–	–	−0. 21	0. 00***	–	–	1.12	0.02**
∆ (Inv)	–	–	1.10	0.00***	–	–	0.10	0.02*
∆ (tot)	–	–	0.01	0. 00***	–	–	−0. 71	0.07**
ECT (−1)	−0. 42	0. 00***	−0. 74	0. 00**	−0. 48	0.00***	−0.64	0.03*

*^,**, and ***denote the 10%, 5% and 1% significance levels, respectively. The Newly–West covariance matrix has been used to correct for autocorrelation and heteroscedasticity observed in the original estimates.

Table 6. Bounds test for post-crisis period

Specification	Selected model	F-Statistic	1%	5%	10%
			I(0) I(1)	I(0) I(1)	I(0) I(1)
F(gdp Debt_n, inf, inv, tot)	ARDL(1, 0, 0, 0, 0)	7. 90	4.94 5.58	3.62 4.16	3.02 3.51
F(gdp Debt_n)	ARDL (1, 0)	6. 14	3.06 4.15	2.39 3. 38	2.08 3.30
F(gdp Debt_g, inf, inv, tot)	ARDL (1, 0, 0, 0, 0)	5. 65	4.94 5.58	3.62 4.16	3.02 3.51
F(gdp Debt_g)	ARDL (1, 0)	5. 54	3.06 4.15	2.39 3. 38	2.08 3.30

Table 7. Long run and short run ARDL estimates (post-crisis)

	F(GDP Debt_n, INF, INV, TOT)		F(GDP Debt_n)		F(GDP Debt_g, INF, INV, TOT)		F(GDP Debt_g)
	Coefficient	P-Value	Coefficient	P-Value	Coefficient	P-Value	Coefficient	P-Value
Panel A: Long-run estimates
Debtg	–	–	–	–	−0. 11	(0.00)***	−0. 05	(0. 02)**
Debtn	−0. 15	(0. 00)***	−0. 03	(0.02)**	–	–	–	–
Inf	−0. 23	(0. 00)***	–	–	−0. 38	(0. 00)***	–	–
Inv	−1. 08	(0. 02)**	–	–	−0. 45	(0. 00)***	–	–
Tot	−0. 15	(0. 00)***	–	–	−0. 00	(0. 00)***	–	–
Panel B: Short-run estimates
∆ (Debtg)	–	–	–	–	0.01	(0. 00)***	1. 38	(0. 01)**
∆ (Debtn)	−0. 44	(0. 00)***	0. 08	(0. 02)* *	–	–	–	–
∆ (Inf)	−0. 31	(0. 00)***	–	–	1.12	(0.00)***	–	–
∆ (Inv)	0.96	–	–	–	0.01	(0. 00)***	–	–
∆ (tot)	−0.11	(0. 00)***	–	–	−0. 71	(0. 02)**	–	–
ECT (−1)	−0. 99	(0. 00)**	−0. 60	(0. 00)***	–	–	−0. 46	(0. 00)***

*, **, and ***denote the 10%, 5% and 1% significance levels, respectively. The Newly–West covariance matrix has been used to correct for autocorrelation and heteroscedasticity observed in the original estimates.

Table 8. Residual diagnostics on estimated regressions

	F(GDP Debt_n, inf, inv, tot)	F(gdp Debt_n)	F(gdp Debt_g, inf, inv, tot)	F(gdp Debt_g)
Panel A: Full-sample
J-B	1.49 (0.47)	0.16 (0.85)	1. 07 (0.58)	1.51 (0.47)
B-G	0.55 (0.58)	0.65 (0.41)	0.31 (0.76)	0.62 (0.51)
ARCH	0.00 (0.96)	0.20 (0.51)	0.39 (0.69)	0.77 (0.40)
White	0.76 (0.51)	0.72 (0.34)	0.82 (0.46)	0.51 (0.76)
Reset	4.42 (0.04)	1.84 (0.09)	1.02 (0.31)	0.99 (0.34)
CUSUM	Stable	Stable	Stable	Stable
CUSUM	Stable	Stable	Stable	Stable
Panel B: Post-crisis
J-B	3.95 (0.14)	3.78 (0.15)	0.75 (0.68)	0.24 (0.89)
B-G	1.27 (0.29)	0.09 (0.89)	0.18 (0.84)	0.06 (0.94)
ARCH	0.55 (0.46)	0.08 (0.91)	0.10 (0.76)	0.07 (0.92)
White	0.53 (0.92)	0.98 (0.36)	0.47 (0.89)	0.98 (0.35)
Reset	2.11(0. 04)	0.99(0. 33)	2.22(0. 03)	1.48(0. 15)
CUSUM	Stable	Stable	Stable	Stable
CUSUM	Stable	Stable	Stable	Stable

p−values are in parentheses ().