I do concede that the software industry can improve on attrition rates and career stability.
But on a simple time value of money front, I'm yet to come across a field of study in the American education system that pays more with lower debt than a STEM degree does.
Skimming the linked paper, it seems to be exclusively discussing PhD programs. Not sure how relevant that is.
https://www.computerworld.com/article/3089314/it-careers/cli...
This isn't the entire issue, I'd agree with that. But it is definitely a relevant piece. And it is notable that tech ceos and politicians are calling this a "shortage" when organizations like RAND are simply seeing a rational, market-based aversion to certain fields.
I'm in the odd position of completely agreeing that the US should have a skills based immigration system like many other countries, but opposing a particular focus on STEM, as I don't like the idea of coercing would-be immigrants into studying this particular field as a condition of coming to the US, and I don't agree that there is any "shortage" that can't be explained as the product of rational decision making about careers.
Americans, by this theory, aren't putting the effort into this difficult field because other grad degrees offer shorter completion times, lower attrition rates, higher pay, and better career stability.
Part of the purpose of the H1b is to create a pool of workers who don't have that freedom of choice, whose right to live and work in the US is controlled by tech corporations.
This is why I, along with a lot of others who would describe themselves as pro-immigration, don't support the H1B program. We're more than ok with talented people coming into the US, but see no reason why we should specifically use our immigration system to increase the number or STEM graduates above. Let people make their decisions according to their interests and market signals. If that means they don't become software developers in Silicon Valley in the numbers tech CEOs think they should, that's the market's answer.
[1]https://www.rand.org/pubs/issue_papers/IP241.html