Suppose we have panel data and estimate once with the "Random Effects Estimator" and once with the "Fixed Effects Estimator". In the second estimation, the coefficients have been greatly reduced, which leads to lower z-statistics if the standard errors remain roughly the same. I wonder if there is an interpretation for this?
The random effects model assumes that the individual-specific effect is unrelated to the other explanatory variables. This is in general difficult to justify in a real data set, because we often just don't know. So I guess the "Fixed Effects Estimator" provides a practical solution and should reflect reality more appropriately. Is the reduction in the coefficient directly linked to that discussion?