There is a cost in doing so, in the

There is a cost in doing so, in the form of a
reduction in the number of available observations. For example, by first
differencing, we reduce the size of the sample by a third. This will, other
things being equal, increase the variance of the resulting estimators. However,
this must be balanced against the possibility that if the intercountry
differences are truly significant, estimates that ignore such differences will be
biased. In the absence of a priori knowledge to the contrary, it is better to
allow for the existence of intercountry differences, even though they may in
fact be negligible, to ensure that there are no significant biases at the cost of
a slight loss of efficiency. After all, the estimators will remain unbiased. We
also find that it is important to use a flexible functional form when
estimating a meta-production function with intercountry data. There will also
be a reduction in the degrees of freedom but there will also be a reduction in
the possibility of bias.