Page 25 - Contributed Paper Session (CPS) - Volume 8

CPS2151 Sarah B. Balagbis
Legend C – two methods comparable; x – proposed method is inferior to the time series cross section
regression (GLS); * - proposed method better than GLS; blank – both values are far from the true value.

4. Discussion and Conclusion
The following conclusions are arrived at based on the simulation results:
1. Covariate parameter estimates change minimally only regardless of the
extent of structural perturbation and/or panel heterogeneity (5% or 10%)
and regardless of the location of the perturbation except when 5% panel
heterogeneity and structural change is spread all throughout the three (3)
locations.
2. Robustness: (a) With no perturbation or with structural change, β1 and β2
estimates are robust and are comparable for the two methods; (b) With
panel heterogeneity, β1 and β2 estimates are better for the proposed
procedure except β2 with 5% panel heterogeneity which proves to be just
comparable; (c) With panel heterogeneity and structural change, β1 and β2
estimates are better for the proposed procedure except when 5%
perturbation are spread all throughout the three (3) locations.
3. Efficiency: (a) With no perturbation or in the presence of structural change,
the covariable parameter estimates for the two methods are comparable;
(b) With panel heterogeneity or a mixture of panel heterogeneity and
structural change, GLS is more efficient though the β1 estimates of the
proposed method can also be considered efficient since the values are
generally small (except when 5% panel heterogeneity and structural change
is spread all throughout the three (3) locations.
4. Reliability: With no perturbation or with structural perturbation, the β1 and
β2 estimates of the proposed method are reliable and comparably reliable
to its GLS counterpart.
Assessing the effect on the ρ estimate when the temporal parameter is
estimated first before the covariate parameters are estimated or when the
temporal parameter is estimated simultaneously with the covariate
parameters would be informative.

14 | I S I W S C 2 0 1 9