The impact of external debt and exchange rate on foreign direct investment in emerging investment markets: new evidence using a PMG-ARDL panel data analysis

Figures & data

Figure 1. Framework of FDI determination.

Source: Author’s estimation.

Figure 2. Representation of data.

Source: Authors estimation.

Table 1. Descriptive statistics.

Criteria	LNFDI	LNExD	LNExR	LNTxR	LNEG	LNINF
Mean	21.85	24.76	4.565	28.387	1.491	2.139
Median	21.782	24.519	4.192	27.292	1.617	2.104
Maximum	24.888	27.059	10.645	34.973	2.730	4.490
Minimum	14.509	22.303	−4.511	19.393	−1.383	0.697
Std. dev.	1.788	1.232	2.989	2.893	0.611	0.706
Skewness	−0.806	0.209	0.112	0.417	−1.522	0.879
Kurtosis	4.19	2.042	3.04	3.32	5.992	4.416

Source: Author’s estimation.

Table 2. CIPS panel unit root test.

Variables	Dimensions	At level	At 1st difference
LNFDI	INTC	−0.858	−8.036*
	INTCT	0.327	−6.809*
LNExR	INTC	−0.65	−7.148*
	INTCT	−2.53*	−5.541*
LNTxR	INTC	−1.088	−6.931*
	INTCT	0.064	−5.570*
LNExD	INTC	2.675	−5.000*
	INTCT	2.57	−4.789*
LNEG	INTC	−2.75*	−5.808*
	INTCT	−1.0723	−3.365*
LNInF	INTC	−2.31*	−10.125*
	INTCT	−3.07*	−8.705*

*1% level of significance.

Note: CIPS = cross sectional Im; Pesaran and Shin W-stat; and INTC = intercept; INTCT = intercept and trend.

Table 3. Result of cross sectional dependency: Pesaran CD (PCD) test.

Variable	Method	Statistics	Probability (P.) value
H0 = No cross-section dependence
LNExD	PCD	5.79	P < 0.01
LNFDI	PCD	17.02	P < 0.01
LNTxR	PCD	15.19	P < 0.01
LNExR	PCD	23.79	P < 0.01
LNEG	PCD	3.16	P < 0.01
LNInF	PCD	6.25	P < 0.01

Source: Author’s estimation.

Table 4. Lag selection criteria.

Lag	LogL	LR	FPE	AIC	SC
0	−1026.59	NA	0.46	16.26	16.40
1	−247.75	1471.84	0.00	4.56	5.503511*
2 (optimum lag)	−211.50	65.07*	3.86*	4.55*	6.31
3	−194.30	29.26	0.00	4.86	7.41
4	−164.75	47.45	0.00	4.96	8.32

Source: Author’s estimation.

Table 5. Result of cointegration test: Kao method.

	t-statistic	Prob
ADF	−3.001	P < 0.01
H0 = There is no cointegration among the variable

Source: Author’s estimation.

Table 6. Robustness check by Johansen Fisher cointegration.

Hypothesized	Fisher stat.*		Fisher stat.*
No. of CE(s)	(from trace test)	Prob.	(from max-eigen test)	Prob.
None	137.8	P < 0.01	84.36	P < 0.01
At most 1	88.7	P < 0.01	61.63	P < 0.01
At most 2	67.91	P < 0.01	50.68	P < 0.01
At most 3	51.53	P < 0.01	43.35	P < 0.01
At most 4	29.63	P < 0.01	26.37	P < 0.01
At most 5	22.51	P < 0.05	22.51	P < 0.05

Source: Author’s estimation.

Table 7. Short-run estimations results of PMG Panel ARDL model.

Variables	Coefficients	t-statistic	P. value
Dependent variable: LNFDI
COINTEQ01	−0.27	−3.33	P < 0.01
D(LNFDI(-1))	−0.01	−0.12	P > 0.10
D(LNEXD)	0.06	0.10	P > 0.10
D(LNEXD(-1))	0.61	1.00	P > 0.10
D(LNEXR)	−0.15	−0.22	P > 0.10
D(LNEXR(-1))	0.43	0.78	P > 0.10
D(LNTXR)	−0.38	−2.56	P < 0.01
D(LNTXR(-1))	−0.07	−0.52	P > 0.10
D(LNEG)	−0.49	−2.27	P < 0.05
D(LNEG(-1))	−0.36	−1.72	P < 0.05
D(LNINF)	−0.08	−0.72	P > 0.10
D(LNINF(-1))	−0.12	−0.67	P > 0.10
C	−1.28	−2.07	P < 0.05

Source: Author’s estimation.

Figure 3. Residual diagnostic: actual and fitted residual.

Source: Author’s estimation.

Table 8. Results of long-run estimations of PMG Panel ARDL model.

Variable	Coefficient	t-statistic	P. value
Dependent variable: LNFDI
LNEXD	0.46	1.55	P > 0.10
LNEXR	0.71	2.53	P < 0.01
LNTXR	0.33	2.39	P < 0.01
LNEG	1.49	3.95	P < 0.01
LNINF	−0.63	−1.78	P < 0.10

Source: Author’s estimation.

Table 9. Pairwise Dumitrescu Hurlin panel causality tests.

Null hypothesis: H0 = the variables (between) does not homogeneously cause	W-stat	Zbar-stat	Prob	Decision
LNEXD ≠ LNFDI	2.84	0.72	0.47	−
LNFDI ≠LNEXD	3.55	1.55	0.12
LNEXR ≠LNFDI	3.29	1.25	0.21	−
LNFDI ≠ LNEXR	3.44	1.42	0.16
LNTXR ≠ LNFDI	2.97	0.88	0.38	−
LNFDI ≠LNTXR	1.52	−0.80	0.42
LNEG≠ LNFDI	2.21	0.00	1.00	−
LNFDI ≠LNEG	2.81	0.69	0.49
LNINF ≠LNFDI	4.28	2.38	0.02	↔
LNFDI ≠LNINF	5.96	4.31	0.00
LNEXR ≠ LNEXD	1.48	−0.84	0.40	←
LNEXD ≠ LNEXR	10.30	9.32	0.00
LNTXR ≠LNEXD	5.38	3.65	0.00	→
LNEXD ≠LNTXR	1.81	−0.47	0.64
LNEG ≠LNEXD	1.84	−0.43	0.67	−
LNEXD ≠ LNEG	2.55	0.39	0.69
LNINF ≠LNEXD	3.49	1.47	0.14	−
LNEXD ≠ LNINF	2.56	0.40	0.69
LNTXR ≠ LNEXR	3.59	1.59	0.11	−
LNEXR ≠LNTXR	1.84	−0.43	0.67
LNEG ≠ LNEXR	1.71	−0.58	0.56	←
LNEXR ≠LNEG	23.70	24.75	0.00
LNINF ≠LNEXR	4.24	2.34	0.02	↔
LNEXR ≠ LNINF	6.70	5.17	0.00
LNEG ≠LNTXR	5.21	3.45	0.00	→
LNTXR ≠ LNEG	2.20	−0.01	0.99
LNINF ≠ LNTXR	2.11	−0.12	0.90	←
LNTXR ≠LNINF	4.51	2.65	0.01
LNINF ≠ LNEG	6.52	4.96	0.00	→
LNEG ≠ LNINF	3.13	1.06	0.29

Source: Author’s estimation.

Note: Null hypothesis statement between the variable denoted by ‘≠’ which indicates ‘does not homogeneously cause’. and unidirectional sign are ← and →, bidirectional sign is ↔ and no causality sign is −.

Data availability statement

Data will be made available based on requests.