This article, and the one it cites, have two serious flaws:
(1) Their "ROI" for college is based on some average of high-school earnings. But average high-school only earnings is the wrong number. We don't care about what 50-65 year-old union employees with a high-school education are getting today, we care about earnings over the next 20 years for today's high-achool, non-college, graduate. That number is dropping, as the number of good paying high-school only jobs decreases. Jobs that a generation ago only required a high-school education now want college (deservedly or not).
(2) The distribution of ROI is based on numbers of institutions, not numbers of graduates. There are far more students graduating from large, cheaper, public schools than Harvard or Princeton, but there are many more small private schools in the distribution of costs. But ROI is not based on a distribution of graduates, it is based on a distribution of schools.
While it does not make sense to take out large loans to finance poor-paying occupations, and it certainly makes more sense to spend less money on college, these articles are very misleading.