Statistical analysis on Cross Country data

Hi all,

Here's a question I have been thinking about:

Often statical analysis is employed on cross country data, where the
authors have tried to gather as much information as possible, for
example on determinants of GNI or the like. Sometimes they manage to
find data on almost all countries. Strictly speaking such a "sample"
is not a real sample, since the sample includes the entire population.
Is there then a need to do statistical analysis, find significance
values and confidence intervals? The mean is the mean and the OLS-
regression coefficients are real. There is no sampling error, just
measurement error. Or what?

And how about the case where the authors find only some of the data?
Or when data from several years is bulked together or spells of years
used as a single dataset? Are they then considered random?

Looking forward to hearing your opinions.

Ole

.