Table 5 Asymmetric coefficient estimates...

Table 5

Asymmetric coefficient estimates

Dep. var: lnGDP		Coefficient		t-statistics
$E C_{t - 1}$		−0.703^***		−3.148 (0.014)
$C G D t_{t - 1}^{+}$		−0.004		−0.369 (0.722)
$C G D t_{t - 1}^{-}$		0.025		1.695 (0.129)
$I P E D_{t - 1}^{+}$		−0.120^*		−1.970 (0.084)
$I P E D_{t - 1}^{-}$		−0.155^**		−2.838 (0.022)
$N F D I_{t - 1}^{+}$		0.158^**		2.707 (0.027)
$N F D I_{t - 1}^{-}$		−0.166^**		−3.316 (0.011)
C		13.119^**		2.744 (0.025)
$Δ \ln G D P_{t - 1}$		0.279		1.371 (0.208)
$Δ C G D t_{t - 1}^{+}$		0.004		0.849 (0.421)
L1		−0.009		−1.119 (0.296)
$Δ C G D t_{t - 1}^{-}$		−0.028^**		−2.612 (0.031)
L1		−0.027^**		−2.682 (0.028)
$Δ I P E D_{t - 1}^{+}$		−0.084^*		−1.883 (0.097)
$Δ I P E D_{t - 1}^{-}$		0.096^***		3.258 (0.012)
L1		0.125^**		2.701 (0.027)
$Δ N F D I_{t - 1}^{+}$		0.170^***		4.476 (0.002)
L1		−0.053^*		−2.280 (0.052)
$Δ N F D I_{t - 1}^{-}$		−0.174^***		−5.739 (0.000)
L1		0.087^*		2.146 (0.064)
Diagnostic tests
$R^{2}$	0.96	$χ_{h e t}^{2}$	15.965 (0.316)
$χ_{s / c o r r}^{2}$	0.041 (0.840)	$J B_{n o r m}$	2.566 (0.277)

Dep. var: lnGDP		Coefficient		t-statistics
$E C_{t - 1}$		−0.703^***		−3.148 (0.014)
$C G D t_{t - 1}^{+}$		−0.004		−0.369 (0.722)
$C G D t_{t - 1}^{-}$		0.025		1.695 (0.129)
$I P E D_{t - 1}^{+}$		−0.120^*		−1.970 (0.084)
$I P E D_{t - 1}^{-}$		−0.155^**		−2.838 (0.022)
$N F D I_{t - 1}^{+}$		0.158^**		2.707 (0.027)
$N F D I_{t - 1}^{-}$		−0.166^**		−3.316 (0.011)
C		13.119^**		2.744 (0.025)
$Δ \ln G D P_{t - 1}$		0.279		1.371 (0.208)
$Δ C G D t_{t - 1}^{+}$		0.004		0.849 (0.421)
L1		−0.009		−1.119 (0.296)
$Δ C G D t_{t - 1}^{-}$		−0.028^**		−2.612 (0.031)
L1		−0.027^**		−2.682 (0.028)
$Δ I P E D_{t - 1}^{+}$		−0.084^*		−1.883 (0.097)
$Δ I P E D_{t - 1}^{-}$		0.096^***		3.258 (0.012)
L1		0.125^**		2.701 (0.027)
$Δ N F D I_{t - 1}^{+}$		0.170^***		4.476 (0.002)
L1		−0.053^*		−2.280 (0.052)
$Δ N F D I_{t - 1}^{-}$		−0.174^***		−5.739 (0.000)
L1		0.087^*		2.146 (0.064)
Diagnostic tests
$R^{2}$	0.96	$χ_{h e t}^{2}$	15.965 (0.316)
$χ_{s / c o r r}^{2}$	0.041 (0.840)	$J B_{n o r m}$	2.566 (0.277)

Note(s): (+) and (−) denote the positive and negative partial sums of the NARDL model, and the p-values are in parenthesis. Diagnostic tests in the model are conducted using $χ_{s / c o r r}^{2}$ for serial correlation, $χ_{h e t}^{2}$ for heteroscedasticity and $J B_{n o r m}$ for normality. The null hypothesis underlying the residual diagnostic tests are as follows: $H_{0}^{s / c o r r} :$ no serial correlation, $H_{0}^{h e t r o} :$ homoscedasticity and $H_{0}^{n o r m} : ε_{t} \sim N (0, σ^{2})$ ⁠. ^*, ^** and ^*** denote statistical significance at the 10, 5 and 1% level, respectively. The estimation of the NARDL model is based on an automatic ARDL framework constructed using optimally chosen SIC- and AIC-based lags

[ViewLarge]