where "it is the disturbance term (i: country, t: time).
One can impose common constant or allow the
intercept term to differ across countries, capturing
country-specific time-invariant effects. This model
is estimated with no intercepts(it¼0), identical
intercept for all pool members(it¼), fixed
effects(it¼i), and random effects(E(i)¼0,
E(i"it)¼0 etc.) techniques (Baltagi 1995). In fixed
effects, the country specific effect is assumed to be
estimable as different intercepts, whereas the random
effects specification treats intercepts as random
variables with zero mean, mutually independent,
homoscedastic, and independent of the disturbance
term. Estimation tries different observation weighting:
(1) to treat all observations with equal weights,
(2) to use generalized least square (GLS) assuming
the presence of cross-section heteroscedasticity,