The easiest solution here is to compare data to countries that allow student loan defaults. AFAIK this is most of countries. There should be good data if the issue is real, or hypothetical, or political lobby based.
Cross-country comparisons like that tend to be hard. It's hard to isolate whether the metrics you're analyzing are because student loan defaults aren't allowed, because of the student selection pressure from an oversized military, because of a culture which views responsibilities and cohesiveness differently, because of secondary funding options, ....
There are many such factors, all of which plausibly matter as much as loan defaults, and since the vast majority of the country you'd like to compare to will tend to be homogeneous with respect to those confounders you often can't find many (if any) actual alternatives to compare to. Statistical approaches to appropriately tease out signal from noise in the larger data set are easy to get wrong, even if you get lucky and actually have comparable data in the first place (e.g., including trade schools or not in the metrics, similarly with associate degrees, counting "in-progress" loans with different payment timelines (treating a truncation bias differently), ...).
Comments
The easiest solution here is to compare data to countries that allow student loan defaults. AFAIK this is most of countries. There should be good data if the issue is real, or hypothetical, or political lobby based.
Cross-country comparisons like that tend to be hard. It's hard to isolate whether the metrics you're analyzing are because student loan defaults aren't allowed, because of the student selection pressure from an oversized military, because of a culture which views responsibilities and cohesiveness differently, because of secondary funding options, ....
There are many such factors, all of which plausibly matter as much as loan defaults, and since the vast majority of the country you'd like to compare to will tend to be homogeneous with respect to those confounders you often can't find many (if any) actual alternatives to compare to. Statistical approaches to appropriately tease out signal from noise in the larger data set are easy to get wrong, even if you get lucky and actually have comparable data in the first place (e.g., including trade schools or not in the metrics, similarly with associate degrees, counting "in-progress" loans with different payment timelines (treating a truncation bias differently), ...).