Page 20 - Contributed Paper Session (CPS) - Volume 8

CPS2151 Sarah B. Balagbis
Specifically, this paper aims to:
1. use the back fitting procedure, forward search algorithm, and
nonparametric bootstrap to the estimation of panel data model with
structural change and panel heterogeneity.
2. assess the robustness, efficiency, reliability and predictive ability of the
estimators through simulation.
3. compare the results of the proposed procedure with the time series
cross section regression estimated using generalized least squares.

2. Methodology
This section presents the procedure to estimate a panel data model with
structural change and/or panel heterogeneity. Estimation will use backfitting
procedure, forward search algorithm, and nonparametric bootstrap
The proposed model without perturbation is given by

In the presence of panel heterogeneity, some cross-sections will follow the
following model:

On the other hand, in the presence of structural change, for some time points
the model becomes:

Furthermore, if both panel heterogeneity and structural change are present,
the model is expressed as

The models assume constant covariate effect (β) across locations and time,
and constant temporal effect (ρ) across locations. Only two covariates (X1, X2)
are considered in this paper. The proposed procedure also assumes the
presence of panel heterogeneity and/or structural change.
The models (1) to (4) are a modification of the spatial-temporal model
proposed by Landagan and Barrios (2007). It only omits the spatial
component. This paper considers only the covariate effect (β) and the
temporal effect (ρ).
To estimate the parameters of the model

given data layout above, the following procedure is proposed:
1. For each t=1,2,…, T in a give panel data realizations { }, i=1,…,n

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