Page 186 - Theoretical and Practical Interpretation of Investment Attractiveness

Calculation strategy. To calculate the panel models described above, the following
steps are required.
First select a free variable. In the regression model, independent variables were
selected based on theory. It must be taken into account that some variables are interrelated .
In cases where the correlation coefficient between two regressors is high (even though
multicollinearity is not an issue) , a high correlation coefficient will increase the standard
errors in the econometric model estimation. When selecting variables for the model, one with
high correlation was selected. Although the simple correlation coefficient does not account
for correlations between time periods or between panel units, it does represent the relationship
between the two variables in question.
Second, the dummy and dependent variables vary across panel units and time.
Intertemporal variation of variables is called within-variance, and variance between regions
is called inter-panel variation and is calculated as follows:
Intertemporal dispersion:
ଶ
ଶ
ଶ
ݏ ௪௜௧௛௜௡ ൌ ଵ σσ ሺݔ െݔҧ ሻ ൌ ଵ σσ ሺݔ െݔҧ ൅ݔҧሻ (4.7)
௜
௧
௜
௜
௧
௜
௜௧
௜௧
ே்ିଵ ே்ିଵ
Dispersion between panels:
ଶ
ݏ ௕௘௧௪௘௘௡ ൌ ଵ σ ሺݔҧ െݔҧሻ ଶ (4.8)
௜
௜
ேିଵ
Total variance:
ݏ ଶ ൌ ଵ σσ ሺݔ െݔҧሻ ଶ (4.9)
௢௩௘௥௔௟௟
ே்ିଵ ௜ ௧ ௜௧
From a mathematical point of view
ଶ
ଶ
ݏ ௢௩௘௥௔௟௟ ൎݏ ଶ ௪௜௧௛௜௡ ൅ݏ ௕௘௧௪௘௘௡ (4.10)
For calculations using panel models, it is desirable to extract the differences between
time and between panel units. In particular, when estimating a fixed effects model, if the
intertemporal variation is less than the variation between panel units, this will lead to
inefficient estimates.
Third , the composite model is estimated using the EKK method as a baseline model.
Although estimating a panel sample using the EKC method is not optimal, when estimating
panel models, it is reasonable to start with the EKC method.
ᇱ
In addition, when calculating a model using the ECC method, ࢟ ൌ࢞ ࢼ൅ ࢻ ൅
࢏࢚
࢏࢚
ઽ compliance with the law of complex error is considered as a necessary condition for testing
ܑܜ
ଶ
hypotheses. ɂ ̱ሺͲǡߪ ሻ This assumption does not hold in panel data, so the estimated
ఌ
୧୲
parameters, although reasonable, are not efficient. For example, because panel data has a time
dimension, these errors are correlated over time. Therefore, when calculating them, it is
necessary to take into account clustering depending on regions, that is, it is advisable to
calculate standard errors corresponding to clustering.
Fourth , although there are several empirical methods based on panel data, the most
common are fixed effects (OLS) and random effects (GLS - Generalized Least Squares, MLE
- Maximum Likelihood Estimation). In this case, if there are strong unobserved intertemporal
invariants (investor culture and customs) and cross-panel variable effects that influence
183