1. Relative rates are fine in aggregate. But the economy doesn't consist of actors who are perfect representations of the median. Individual companies are going to be net beneficients or net losers. And it is always the net losers who are more vocal, complaining, and ultimately influential over the wider market. Actor confidence in large markets is naturally biased towards fear, not security. This is one of the reasons why regulation is sometimes welcomed by actors, if they can influence all actors equally.

2. "Treat people nicely" is about as effective a corporate policy as a law is forcing people to "be good and moral." A small company can effectively assert hiring control in hiring only people who "gel" with the people leading the company. A large company with a hiring quota of hundreds or thousands of people a quarter cannot. Ultimately, people are going to have personality conflicts with coworkers at large companies. You can either accept that as an inevitability and find a solution for it, or you can write off people finding (somewhat legitimate but still entirely foreseeable and unavoidable) excuses to move as BigCo management not treating their workers "nicely".

So the only alternative is to let corporations offload their training externalities onto higher education?

If that's what you're saying, it's a false dichotomy.

Cry me a river for the corporations, who cares what they want? It's always their agenda that wins these days anyway. That's the way the world is going.

No, what I'm saying is that we need to find some kind of way that balances protecting workers from exploitation with the kind of security that allows corporations to invest in workers' training.

Yes, one way, in theory, of doing that is by offloading training externalities to higher education. But that has its own tradeoffs, namely, the tendency of universities to adopt ivory-tower attitudes to education, the tendency of American higher education to inflate costs beyond any reasonable limit to afford facilities, services, and administration of dubious value to the education afforded the end consumer / student, and the tendency of universities to not educate with an eye to the skillsets which contemporary employers value.

Which is why the better way is to localize training efforts with the actors for whom it is most relevant. But corporate actors are not going to do "the right thing" in a vacuum, the law needs to empower them to do "the right thing", because typically "the right thing" will hurt any actor who does it in isolation but is bearable if all actors commit to doing it together. The easiest, safest, and most predictable way of making that happen is through law or regulation.

Regarding #1: From the point of view of the company, if other companies are educating their employees, then the company will likely stop spending the resources on their own employees while still taking advantage of the workers educated at the other companies. So, this point doesn't make sense without a very optimistic view on the cooperation that would take place between companies.

2. Treating 'nicely' is far from enough if another company is simply able to offer a better package. For example, are small/new tech companies not deserving of keeping their investments because they can't pay their devs as well as some large ones? Or if they can't match the prestige of other companies?

The problem with #2 is that if a company spends $30k of time and effort to get an employee trained up, the poaching company can offer a raise of 10k and still be in the black 3 years later. If the original company decided to match that offer, they're now spending 60k more over three years, and the poaching company can offer a 15k raise and STILL be in the black after 3 years. Any money you pay in training is a sunk cost that doesn't have to be paid by your competitors.

Sure but presumably, unless all employees start out only at your company, then the effect would also sometimes happen in your company's favour.

As I said in point #1, as long as the rate of defections/arrivals is about even, then your losses on one employee would be offset by your gains on others.

Having an even defection/arrival rate requires other companies to also train their employees, and not only hire previously trained persons.

What makes you think the poached employee is actually going to stick around for 3 years, instead of jumping ship when Company C comes along and poaches them for a $10k raise? Most people in Silicon Valley these days don't stay in jobs for 3 years, according to what I've read.