Economic policy uncertainty and stock returns among OPEC members: evidence from feasible quasi-generalized least squares

Oyewole, Oluwatomisin J.; Adubiagbe, Idowu A.; Adekoya, Oluwasegun B.

doi:10.1186/s43093-022-00124-w

Future Business Journal

Table 1 Preliminary test results

From: Economic policy uncertainty and stock returns among OPEC members: evidence from feasible quasi-generalized least squares

Country	Start date	Summary statistics				Unit root test		Persistence and endogeneity tests
Country	Start date	Mean	SD	Skewness	Kurtosis	Level	First difference	Persistence	Endogeneity
EPU	7-Jan	134.288	45.919	0.89	3.777	− 3.744^b***	–	–	–
Stock returns
Algeria	7-Jan	1.012	0.12	11.061	129.188	− 12.78^a***	–	0.844***	0.024
Ecuador	12-Feb	1.004	0.021	0.313	3.406	− 9.104^a***	–	0.804***	− 0.021*
Iran	7-Jan	1.078	0.752	11.526	136.747	− 10.77^a***	–	0.844***	− 0.066
Iraq	9-Sep	1.099	1.12	10.347	108.39	− 9.962^b***	–	0.791***	− 0.102
Kuwait	7-Jan	0.996	0.051	− 0.905	6.416	− 6.117^a***	–	0.787***	− 0.024
Nigeria	7-Jan	1.002	0.075	0.391	8.94	− 5.771^a***	–	0.844***	− 0.067**
S. Arabia	7-Jan	1.002	0.0671	− 0.25	4.724	− 10.75^b***	–	0.844***	− 0.038
UAE	7-Jan	1.005	0.055	− 0.035	4.83	− 6.262^b***	–	0.844***	0.011
Venezuela	7-Jan	1.163	0.486	4.546	29.7	− 1.74^a*	–	0.844***	− 0.075

The null hypothesis for the autocorrelation is there is no serial dependence, while for heteroscedasticity is that there is no conditional heteroscedasticity. The persistence test is conducted by regressing a first-order autoregressive process for the predictor e.g. \(z_{t} = \varphi + \sigma z_{t - 1} + v_{t}\) using OLS estimator. The first order autocorrelation coefficient (σ) captures the persistence effect and is reported in Table 1. The null hypothesis is that there is no persistence effect in the predictor, while the alternative is that there is persistence effect in the predictor. For endogeneity, it involves a three-step procedure. Firstly, we run the following predictive model:: \(s_{t} = \alpha + \beta g_{t - 1} + \varepsilon_{t}\), where \(s_{t}\) is the stock returns and \(g_{t - 1}\) the natural log of global economic policy uncertainty. The second step is that we follow the Westerlund and Narayan [26] by modeling the predictor as follows: \(g_{t}\) = \(\mu \left( {1 - \rho } \right) + \rho g_{t - 1} + v_{t}\) and in the final step, the nexus between the error terms is captured using the following regression: \(\varepsilon_{t} = \gamma v_{t} + \pi_{t}\). If the coefficient \(\gamma\) is statistically different from zero at any of the conventional levels of significance at 1%, 5%, and 10% respectively, then the predictor variable is endogenous; otherwise, it is not. a and b respectively denote models with intercept only and intercept and trend. ***, ** and * represent significance at 1%, 5% and 10% critical levels, respectively.

Back to article page