Dynamic panel data model estimation results for equation 8 in model 2 are presented in appendix 5 to 8 and reported in Table 5.2. The system GMM estimator is categorized into the one-step and two-step option, and is reported in columns 2 and 3 respectively. The OLS estimates and the LSDV estimates are reported in columns 1 and 4 respectively. All estimations are again in this case robust to heteroskedasticity or autocorrelation. This is irrespective of the option under which the estimates are considered. Details of these results are in appendices 7 to 11.
The related predetermined and endogenous variables on the right hand side of this specification include the lagged INV and WR respectively. To control for endogeneity of these variables that appear as regressors, internal instruments are again utilized; and these include the lagged levels of the standard differenced equation (equation 10) and lagged differences of the levels equation (equation 8). The list of these internal instruments are included in the results output and can be found in appendices 7 and 9. Correlation coefficients (see appendices 8 and 10) between residuals from the base regression and independent variables were computed as an additional check of potential
123 endogeneity problems. An investigation of these coefficients of correlations suggests that none of the independent variables is highly correlated with predicted residuals.
Table 5.2: Estimated Empirical Results in the Investment-Remittances Model (Model 2) Dependent Variable: INV
OLS SYSTEM-GMM LSDV
One-Step Two-Step
Instrument Weight collapsed collapsed
Regressors (1) (2) (3) (4)
INV(-1) 1.49215* 1.59850* 1.52204* 1.66606*
(0.067) (0.079) (0.079) (0.112)
WR -2.04091* -2.21179* -2.09074* -2.19928*
(0.103) (0.124) (0.135) (0.093)
INTR -17.550 -22.14462 -4.80923 3.64175
(11.838) (26.458) (22.0004) (10.959)
INF 3.81952 -1.51909 0.95473 3.78888
(6.265) (4.505) (3.862) (5.518)
FD 1.07333* 1.33819** 0.87714*** -0.24794
(0.262) (0.639) (0.484) (0.299)
Constant -225.1606 -131.8035 -70.604 -476.703
(106.741) (162.038) (178.967) (150.954)
Time Dummy Yes Yes Yes Yes
Country Dummy No No No Yes
Observations 147 147 147 147
No. of countries 21 21 21 21
Instrument count - 13 13 -
F-stat (Wald χ2 ) 95.80 2378 260.57 108.82
F-stat (p-value) [0.0000] [0.0000] [0.0000] [0.0000]
AR(2) - [0.146] [0.196] -
AR(3) - [0.127] [0.225] -
Sargan Test (OIR) - [0.000] [0.000] -
Hansen Test (OIR) - [0.262] [0.262] -
Notes: Robust standard errors, consistent in the presence of any pattern of heteroskedasticity and autocorrelation within panels are reported in curly brackets.
Robust standard errors are with Windmeijer (2005) finite-sample correction for the two-step covariance matrix
P-values are reported in square brackets * indicates significant at 1 percent level ** indicates significant at 5 percent level *** indicates significant at 10 percent level
124 Some specification tests are again examined as a starting point to determine the reliability of coefficient estimates reported in Table 5.2. The working assumption that the idiosyncratic errors in the system GMM estimators are serially uncorrelated for consistent estimations is retained. The assumption that the full disturbance is autocorrelated because it contains fixed effects is not relaxed hence; the system GMM estimator remains the most appropriate tool to eliminate this source of trouble. The Arellano-Bond test for autocorrelation are reported as AR(2) and AR(3) in the lower portion of table 5.2. The p-values are greater than 0.05 in the one-step and two-step system GMM estimates indicating that there is no evidence of serial correlation at the five percent level of significance. Given these results, the estimates can be regarded as consistent and the instruments are not endogenous.
The test of overidentifying restrictions of whether the instruments, as a group, appear exogenous is implemented by the Sargan and Hansen J tests. Here the Hansen J statistic, which is the minimized value of the two-step system GMM criterion function, and is robust to autocorrelation, is of tremendous importance. Only the respective p-values are reported for this test results in the lower part of table 5.2. of course, the null hypothesis that the population moment condition is valid is not rejected if In columns 2 and 3, the summary statistics indicate that the system dynamic panel model of the selected 21 SSA countries has 13 instruments and 12 parameters in both the one-step and two-step system GMM options. This represents a total of 1 overidentifying restrictions in each case. Thus, the Hansen–J statistic does not reject the Over-Identifying Restrictions (OIR), thus confirming that the instrument set can be considered valid. The F-statistic which measures the overall significance of all regressors in the estimated model is satisfactorily significant at the one percent level. This of course is indicative of the fact that
125 all the exogenous variables, in each estimated result, jointly explained significantly, the systematic variations in domestic investment across the sampled SSA countries over the study period.
A look at the control variables reveals the coefficient estimates are sufficiently consistent with theoretical expectations. The Blundell–Bond robust estimates of lagged domestic investment are positively signed. As can be seen in columns 2 and 3 of table 5.2, past realizations of domestic investment positively impact on its contemporaneous levels. These domestic investment dynamics are significant at the 1 percent level in these specifications. In specific terms, a 10 percent increase in domestic investment dynamics will explain about 15.98 percent and 15.22 percent of the increase in the contemporaneous realizations of domestic investment within the sampled SSA countries. This of course is when the one-step and two-step collapsed instruments options are considered respectively. Domestic investment dynamics suggest here that domestic investment in SSA has a way of feeding on its past values.
Workers‘ remittance variable also has highly significant results in both the one-step and two-step system GMM robust estimates. In all, this set of results for both specifications produced a contemporaneous negative impact on domestic investment across the sampled countries over the study period. The levels of significance here are all one percent. In more definitive terms, a 10 percent increase in workers‘ remittances under the Blundell–Bond estimates, will explain negatively about 20.9 percent, of the changes in domestic investment across the study group. This is suggestive of the possibility of a crowding-out of domestic investment role for remittances in the selected SSA economies. The negative nature of this relationship is suggestive of the fact
126 that remittances flow to SSA is basically a financial flow and does not necessarily double as capital flows.
Financial deepening variable has some significant results ranging from the five percent to the ten percent levels. Given the collapse option in the one-step and two-step system GMM, a 10 percent increase in financial deepening will produce about 13.38 percent and 8.77 percent increase in domestic investment respectively. Given this finding, it can be remarked here that policies which encourage banks to increase provision of financial services that have wider choice of services geared to all levels of society will help attract more remittances to Africa. Remittances related banking products or services will be immensely useful in this regards.