What effect do you think that has?

If smart students choose selective schools and smartness is associated with ROI, we would expect to see a correlation between ROI and school selectivity. That we don't is interesting.

If merit is highly correlated with school selectivity, then selectivity can be considered a proxy for smartness/determination/etc. If that's the case, education ROI appears to be not strongly correlated with merit (via the selectivity proxy). That is very interesting.

Actually, the chart shows a pretty damning story for selection bias: despite all the selectivity of schools, there's no sign of improvement in returns for more selective engineering schools. It makes a bit of intuitive sense: any MIT/CMU/Stanford/etc premium on degrees would be capped by management salary structure, since fresh grads don't typically earn more than their boss. It's possible that a linear fit is not a great model, as the little data there is on the far left chart is all above average, but the 10x engineer isn't registering on the radar.

As for what people can do differently: if you have equal chances of succeeding at the most selective arts and humanities college, or the least selective engineering, choosing engineering will net you better 20 year salaries. Especially relevant if you're considering humanities as a law degree path -- the LSAT has logical reasoning test, and if you can hack that, you can probably do okay in engineering math / CS classes.

This is so true. I am not sure we can accept or reject any hypothesis unless the treatment factors were assigned in a completely randomized way.