“…We were, in fact, able to obtain even opposite coefficient signs in some of our regressions presented in the next section -just by a small changes in the number of lags used as instruments, and the specification tests did not provide us with almost any guidance to select which of the results are more relevant than others. Let us note that the results from our GMM experiments are available upon request, although not worth a red cent 10. The bootstrap corrected fixed effect model was estimated in Stata, using the xtbfce command (De Vos,Everaert, and Ruyssen, 2015).…”