A note on cross-country growth regressions