Dynamic panel data model estimation results for equation 13 in model 3 are reported in Table 5.3. The results of the Blundell–Bond system GMM estimator as in previous cases are reported in columns 2 and 3 of the table. In addition to the system GMM estimator, the OLS and the LSDV estimators are also reported in columns 1 and 4 of the table respectively. The complete results are presented in appendices 9 to 12. As usual, all estimations are robust to heteroskedasticity and autocorrelation. Details of these results are in appendices 9 to 12.
The related predetermined and endogenous variables on the right hand side of this specification include the lagged REB on the one hand, and REER, CAB as well as WR on the other hand respectively. As in previous two cases, I control
127 for endogeneity of these variables that appear as regressors by utilizing internal instruments that include the lagged levels of the standard differenced equation and lagged differences of the levels equation. The list of these internal instruments can be found in appendices 14 and 16. Correlation coefficients (see appendices 15 and 17) between residuals from the base regression and independent variables were computed as an additional check of potential endogeneity problems. An investigation of these coefficients of correlations suggests that none of the independent variables is highly correlated with predicted residuals.
128 Table 5.3: Estimated Results in the Trade Balance-Remittances Model (Model 3) Dependent Variable: REB
OLS SYSTEM-GMM LSDV
One-Step Two-Step
Instrument Weight collapsed collapsed
Regressors (1) (2) (3) (4)
REB(-1) 0.54847* 0.32606* 0.33445* 0.36634*
(0.093) (0.055) (0.056) (0.093)
WR -0.44741* -0.20740** -0.22063* -0.22396**
(0.101) (0.074) (0.072) (0.102)
CAB 0.60073* 0.63834* 0.63423* 0.65873*
(0.072) (0.019) (0.026) (0.064)
INTR -22.0679*** -20.4288 -9.8458 7.78239
(11.175) (11.842) (10.463) (8.989)
OPEN -13.9869 54.7389 -82.0211 147.7737
(69.580) (177.83) (131.248) (265.901)
REER 0.07964 0.090 0.03815 -0.60125
(0.124) (0.210) (0.1548) (0.433)
Constant 57.036 -301.372 -184.3704 -171.851
(124.148) (291.56) (273.769) (169.054)
Time Dummy Yes Yes Yes Yes
Country Dummy No No No Yes
Observations 146 146 146 146
No. of countries 21 21 21 21
Instrument count - 19 19 -
F-stat (Wald χ2 ) 135.08 5881.31 3994.26 60.27 F-stat (p-values) [0.0000] [0.0000] [0.0000] [0.0000]
AR(2) - [0.277] [0.378] -
AR(3) - [0.146] [0.267] -
Sargan Test (OIR) - [0.000] [0.000] -
Hansen Test (OIR) - [0.219] [0.219] -
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
Relevant specification tests results are first examined in Table 5.3. The working assumption that the idiosyncratic errors are serially uncorrelated
129 for consistent estimations under the system GMM estimators is still retained.
The Arellano-Bond test for autocorrelation AR(2) and AR(3), are reported in the lower portion of Table 5.3. Interestingly from these results, there is no evidence of serial correlation at the five percent level of significance. This provides sufficient basis to conclude that the coefficient estimates can be regarded as consistent. The p-values of the Hansen J tests statistic indicate that the system dynamic panel of the selected 21 SSA countries has 19 instruments and 13 parameters. This represents a total of 6 overidentifying restrictions in each of the options in columns 2 and 3 of table 5.3. Consequently, 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 is highly significant at the one percent level. This result is indicative of the fact that all the regressors jointly explained significantly, the systematic variations in real external balance (REB) across the sampled SSA countries over the study period.
The control variables in the estimated results are next considered. These figures reveal some impressive and striking results which are largely significant and sufficiently consistent with theoretical expectations. The Blundell–Bond two-step system GMM robust estimate of lagged real external balance is positively signed and significant at the one percent level. As can be seen, this result indicates that past realizations of real external balance positively impact on its contemporaneous levels. In specific terms, a 10 percent increase in real external balance dynamics will explain about 3.34 percent increase in the contemporaneous realizations of real external balance within the sampled SSA countries. The applicable level of significance here is one percent. And the collapsed instruments option was utilized.
130 Workers‘ remittances variable produced a highly significant result in the Blundell–Bond system GMM two-step robust estimates. A particularly striking thing about this result is that it is negatively signed and significant at the one percent level. This clearly suggests that workers‘ remittances inflow depresses foreign trade balance in SSA. Contemporaneously, real external balance in the selected SSA countries decline by about 2.21 percent as workers‘ remittance inflows into SSA rise by 10 percent. This finding suggests that the bulk receipts from workers‘ remittances flows to SSA may actually be channeled into the consumption of imported goods. This gives the impression that workers‘ remittances flows are potentially harmful to the economies of the receiving SSA countries if deliberate policies to channel such flows into productive uses are not formulated and enforced.
Current account balance (CAB) is another variable that is positively signed and highly significant at the one percent level. Interestingly, this result is not unexpected since a positive CAB is largely indicative of healthy domestic economy which provides a platform for a favourable trade balance. In more definitive terms, real external balance will increase by about 6.34 percent for every 10 percent rise in current account balance. By implication, policies aimed at boosting the current account balance of SSA economies will also assist in boosting the trade balance of these economies.
The interest rate variable is negatively signed and insignificant even at the ten percent level. The meaning of this is that every rise in the domestic interest (lending) rate in the SSA economies increases the cost of production and by implication reduces competitiveness of all exported goods. The direct consequence of this will be a depressed trade balance for these SSA economies. Trade openness and real effective exchange rate variables are also insignificant explanatory factors of the changes in real external balance. This is
131 not surprising given the very low size of the external sector and the preponderance of primary products in the total exports of most SSA economies. However, the negative signs of the openness variable in two out of the four specifications call for some form of guided deregulation in SSA economies seeking to completely open up to the rest of the world economies.
A quick remark that must be made regarding the results in the three estimated models is the failure of the system GMM coefficient estimates in some cases, to lie within the boundaries created by the OLS and LSDV estimates as prescribed by Bond, Hoeffler and Temple (2001). In checking for the source of this discrepancy, the validity of all instruments used in the model estimation was examined and found to be satisfactory. The only logical explanation for the failure of the Bond, Hoeffler and Temple (2001) simulation prescription to apply to the analysis in this work is the possibility that the simulation exercise is very likely sensitive to data employed. This openly calls to question the validity of the position of Bond, Hoeffler and Temple (2001) regarding the boundaries created by OLS and LSDV estimates for the system GMM estimates.