>38,000 people on a planet of 6 billion is way too small a sample size.
>Just how did Pew come up with those 38k people anyway. Underdetermination
>problems are immense on polls like that.
It was 44 countries, not the entire planet. And a sample of slmost 1000 per country isn't small - a routine Gallup poll of the U.S., a populous country, uses about 1000 people. It's not going to give you second-decimal-place precision, but it's good enough to determine trends. And these margins in this survey were quite large, well beyond any standard error.
Besides, the findings are seconded by the repeated electoral failures of protectionist parties and candidates. I'm more convinced than ever that this whole "globalization" issue is badly misspecified. It's a good thing too, because I've got a book coming out with a long chapter making that argument.
Doug