ABSTRACT
In Chapter 9 we introduced panel data by showing how we could extend both our macro and micro production function using two waves to control for unobservables which were time-invariant. With two waves of the data we could estimate a pooled equation, an equation with a dummy capturing the time-invariant unobserved and a differenced equation. Now clearly we do not need to stop at two years of data. So in this chapter we present the alternative estimators that are available to us with a full panel. These estimators are the fixed effect (FE), the first difference (FD), the pooled OLS (POLS) and the Random Effects (RE) and they are the subject of Section 10.2. We then present some methods for testing for the nature of the unobservables in the data in Section 10.3. As we see in Section 10.4, the choice between the estimators depends on the nature of the endogeneity problem in the equations. In Section 10.5 we compare the results across the alternative estimators. We then extend our analysis of the Ghana firm panel data of the last chapter.
