K. Ludi: consumption behaviour in Zambia
The actual coefficients for the regression of equation (1) can be found in table 2.
Table 2: Regression results of the cointegration equation (dependent variable: LN_CONS_ZK)
Sample (adjusted): 1971 2001 Included observations: 31 after adjustments | |||||
Variable |
Coefficient |
Std. Error |
t-Statistic |
Consistent with |
Interpretation8 |
LN_GDP_ZK |
0.999958 |
0.001639 |
610.0092 |
Yes |
An increase in national income will |
R_LENDRATE |
-0.002285 |
0.000585 |
-3.903330 |
Yes |
An increase in short-term lending |
DUM_92 |
-0.309545 |
0.038157 |
-8.112476 |
Yes |
The dramatic decline in PCE in 1992, |
R-squared 0.713131 Adjusted R-squared 0.692640 |
An Augmented Dickey Fuller (ADF) unit root test was performed on the residuals of the
cointegration equation, and it was concluded that the residuals are indeed stationary.9
4.2. The Error Correction Model (ECM)
The ECM represents the short-term relationship between the data series, and is based on
the proposition that a proportion of any disequilibrium from one period can be corrected
in the next period. Long-run effects are included in the short-run equation via the lagged
residuals of the cointegrating equation (res_coint).
Attempts to estimate equation (2) as the ideal short-term situation were statistically
fruitless. Thus, other short-run factors were included to be statistically successful and
Due to the non-stationarity of the variables, the magnitudes of the coefficients cannot be
stringently evaluated, and the variables may not be statistically evaluated.
See Appendix B for the unit root tests of the cointegrating residuals.