The study you linked is over 10 years old. Furthermore, it is using lifetime earnings as the core metric, which means they are pulling in data about people who earned their degrees like 50 years ago.
All the data that currently exists shows better outcomes for students that go to college.
One would expect this even if college had no benefit to students because the population of students that go to college is pre-selected. Before they attend, students that go to college, on average, demonstrate better analytic skills than the students that do not go to college. They also, on average, have access to more existing wealth and other resources through their family.
In the absence of perfect data (which is almost always the case in sociology), it is reasonable to look at case studies to try to make sense of reality. It is not bad practice. It is what Harvard Business School does. It's what product managers and UX designers do when creating products. It's what marketing teams do when selling products.
Feel free to disagree with the interpretation of anecdotal data, but statements should not be dismissed out-of-hand because no p-value accompanies them.
All the data that currently exists shows better outcomes for students that go to college. One would expect this even if college had no benefit to students because the population of students that go to college is pre-selected. Before they attend, students that go to college, on average, demonstrate better analytic skills than the students that do not go to college. They also, on average, have access to more existing wealth and other resources through their family.
In the absence of perfect data (which is almost always the case in sociology), it is reasonable to look at case studies to try to make sense of reality. It is not bad practice. It is what Harvard Business School does. It's what product managers and UX designers do when creating products. It's what marketing teams do when selling products.
Feel free to disagree with the interpretation of anecdotal data, but statements should not be dismissed out-of-hand because no p-value accompanies them.