Page 262 - Contributed Paper Session (CPS) - Volume 2
P. 262
CPS1853 M. Irsyad Ilham
Number of motor vehicles signed as unit. The source of data is Indonesia
Statistical Agency and Ministry of Environment and Forest. Panel data was
combination form of time series and cross section data. The combination of
time series and cross section data was used to prevent such lacking of time
series and cross section data itself. Moreover, panel data were capable in order
to answer the questions which time series or cross section data cannot
answered. Besides, some advantages in using panel data according to Baltagi
(2005):
1. Panel data would controlling about individual heterogeneity;
2. Panel data give more informative data, more variability, less collinearity
among the variables, more degrees of freedom, and more efficiency;
3. Panel data are better able to study the dynamics of adjustment;
4. Panel data are better able to identify and measure effects that are
simply not detectable in pure cross section or pure time series data;
and
5. Panel data models allow us to construct and test more complicated
behavioral models than purely cross section or time series data.
To determine the most appropriate model used in the study conducted
model significance test. Model selection can be done informally or formally.
According to Gujarati and Porter (2008) there are four considerations that can
be used to choose the best model between fixed effect or random effect
model, namely:
- If the amount of time series (T) data is large and the number of cross
section (N) data is small, the difference between the fixed effect and
the random effect model is very small, so the choice is based on the
ease of calculation, ie the fixed effect model.
- When the amount of time series (T) data is small and the number of
large cross section (N) data, the estimates obtained by the two
methods can differ significantly. In the random effect model, α_i = "" α
+ μ_i where μ_i is the component of individual error and αi in fixed
effect model is not random. If the individual or unit of the cross section
of the sample used is not random, then the fixed effect model is more
appropriate to use. Whereas, if the cross section unit is random, the
random effect model is more appropriate to use.
- If the individual error components μ_i and one or more regressor are
correlated, the estimator derived from the random effect model is
biased, whereas the fixed effect model is unbiased so that the fixed
effect model is better used.
- If the number of large cross section (N) data and the number of small
time series (T) data and the assumption of the random effect model
are met, the random effect model estimator is more efficient than the
fixed effect model estimator.
251 | I S I W S C 2 0 1 9