This is like The Lord Kaldor analytical scandal of the 1960s when the professor was selective to prove his thesis about productivity in rapidly growing economies, i.e. Verdoorn's Law. If one took Japan out of the data set the equations were insignificant as Prof. Bob Rowland was keen to point out.
Econometric analysis that is cross sectional is always fraught with risk that the numbers of outliers will distort a thesis. Eg Japan in a world of low performers.
Many economists were refused PhDs because they included evidence that did not actively support a theory within a thesis. I had that very issue myself, but my honesty prevailed. It was just as important to show the theory was not generalizable. In this way we see that apart from accounting models, generalisations of economic laws fall apart across countries and across time. Tell that to Lord Jevons and his economic cycles arising from sunspot cycles!
Sad to see research that is discredited as a generalisation, but the fact remains that there are very few generalisations that can be made in economics apart from the famous consumption function of Lord Keynes that has stood the test of time in industrial economies. Apply that model to developing economies and the laws fall apart. Horace's for courses!
Don't generalise, just research!