17 The constant coefficient model with residual homogeneity and normality can be estimated by using pooled
Ordinary Least Square (OLS) regression model, which refers to constant intercepts and slopes across time and
regions. More specifically, there is assumed to be unobserved individual heterogeneity. Since the independence
and homoscedasticity assumptions of the error terms in the pooled OLS are not usual in the panel data
applications, it will not be so realistic to expect that pooled OLS be adequate for such models (Davidson and
Mackinnon, 1993).
18 A main assumption in random effects model is that the random effects and explanatory variables are not
correlated with each other. For this reason, a Hausman test is employed to make a specific comparison between
the fixed and random effects estimates of coefficients (Baltagi, 2001). The results of the Hausman test indicate a
Chi-square statistic equal to 8.843 (with 3 degrees of freedom) for cost efficiency and 8.771 (with 3 degrees of
freedom) for profit efficiency, which are both statistically significant at the 10% level.
19 Generally, not all investors tend to keep their stocks in the long term and benefit from dividend payments.
However, these groups of investors will be interested in the stream of expected future dividend payments since
stock price will reflect the present value of future dividends. Therefore, positive changes in the expectations
about the dividend payments will be reflected positively in the stock returns, providing the investors to earn
profits on sale (Board and Day, 1989).