Similar to the discrepancy between "normal" and "actual" bilateral trade, one may ask the question about the difference between
"normal" and actual bilateral multinational activity. However, with multinational activity, zero bilateral data and heteroskedasticity
are very important, even more so than with trade data. Therefore, this paper suggests using generalized linear rather than
log-linear models to specify "normal" FDI and obtain estimates of unexhausted FDI potentials. I use panel data on Austria's
bilateral multinational activity across 25 countries and 7 country-blocs, 4 sectors and 13 years to illustrate the disadvantage
of log-linear model estimation at quasi-maximum likelihood estimation.
