If you’re talking about the “elite” schools - Ivy or otherwise - there’s a little bit more to it.
A resume is a really, really low bandwidth way to get a feel for someone. Of that’s all you have to go on for starters, it at least tells you which gauntlets they’ve already run. It’s like hiring someone who has worked at Apple or Google for ten years.
As a simplifying assumption, think of ability as a normal distribution - a bell curve. The average on Stanford grads may be higher than those of Liberty University, although there still may be enough overlap that you can’t say that any given candidate is better from one or the other.
If you’re talking about someone who transferred out of Harvard to go to Austin University or whatever they’re calling themselves, that opens up an entirely different set of questions.
No worries about the lack of sleep. I’ve been there and then some.
I do think however that you’re misinterpreting my argument to at least some extent.
First, it’s a completely noisy signal. It’s also, unfortunately, the only thing we have when a CV lands on our desk. It obviously decreases in importance as the number of positions held/publications made/other experiences increase. If someone were to have a dozen pubs in reputable journals and ten years experience working in what I’m interested in, I’m not going to take their school into account. The other, later work is much more relevant. If on the other hand they transferred from MIT to Liberty University and that’s the only data point I have, that’s what I am going to need to go off of. I have a lot of resumes to look at, and still have to do my regular full time job. I’m not arguing that it’s not noisy. I’m just pointing out that if we consider something like a weighted function in CV evaluation, the fewer items there are, the higher the weights assigned to non-preferred variables might be. I’ve collaborated with researchers from some of the most respected institutions in the world, and other than arrogance I can’t say that they had a whole lot in common.
Second, I do not think ability falls on a bell curve. I believe talent is a highly skewed distribution. It might get more normal the more you remove sources of variability - I don’t think you could pick someone at random off the street and ask them to write up a Bayesian classifier, but if you reduced the sample down to stats/ML grads, you’d probably find some are better and some are worse but you might see a meaningful average being drawn. I was just trying to make it easier to visualize. I am an actual data scientist (well, complexity theorist), and I am not naive about data.
In terms of social power, that’s absolutely one of the main reasons people pay the outrageous tuitions for those institutions. I do need to note for anyone reading along that those same institutions will waive tuition if your family income is below $150k or so, so do not write them off. We need more diversity.