Wednesday, January 13, 2016

Maybe...But Not Necessarily

I suppose this should get filed under problems with statistics.  This Vox article is talking about maps showing student loan balances and delinquencies (which is sourced from the Washington Center for Equitable Growth), and while it may seem that their conclusion is logical, it may not be factual.

They are stating that since the high delinquencies map is practically the inverse of the high loan balances that people with lots of student loans are doing ok while those who are having problems don't have high balances, and that seems to follow but maybe not.

The problem is the way averages work.  If one area has lots of college grads with student loans and most of them are doing well, then it would follow that you get low delinquency rates, but also that the loan balances would be higher (high density college grads would imply greater need for higher education to live/work in the area, which would mean more and more expensive education).  On the other hand places with lots of college dropouts and with people not generally doing well, are likely to have lower debt and higher delinquencies.  BUT without knowing the delinquency rates and the distribution of the loan balances you can't say anything about the relationship.

What you want to see is a plot of delinquency rates at various loan balance ranges, and maybe one showing loan balances vs. income level.  This doesn't mean the drawn conclusion is wrong, but just that it shouldn't have been drawn in the first place.

No comments: