Why don't tech companies pay their engineers to stay?

This is like saying, "Making people drive slower will reduce total number of accidents. So lets just reduce all speed limits to 25 miles an hour. It's only the obvious solution right?".

This kind of lazy thinking is called incompetence, because finding really good balanced solutions is actually work. This work involves observing problems, understanding root causes, piloting and experimenting, bringing other mangers, skips and CXOs into alignment, executing retention deftly while keeping reports happy. But this is too much work for managers (but of course too much work for engineers is never too much work).

What a joke! No wonder SWEs complain so much about management.

The last company I worked for gave me significant golden handcuffs. Probably 30-40% more than I could get elsewhere. They had a "cost of replacement player" policy, where they count the cost of hiring, sourcing, training over the years, etc, and come up with a number thats almost always way above market rate for people who have been around a bit.

Attrition was lower than industry average...but not by much. The reality is that in big tech, salaries are high enough that people stop caring. People left for a variety of reasons, the most common one being "I want to see how it is elsewhere" (especially from college hires who have been at the company several years). Others leave because they want to work at a startup. Some leave to join their friends. Some because they move to follow a significant other who got a job elsewhere (many prefer offices over WFH, so offering that only helps so much).

In the end, again, retention is higher, but when you crunch the numbers, its very debatable if it's worth it for the company. Worse, sometimes companies end up "competing" over it, so even if you pay a lot, the next one offers more because they know your salary bracket, then you raise yours until one gives. That sounds great for the employee but there are budgets and limits at some point.

And then there's the topic of counter offers. The uncomfortable truth is that there's very significant demographic differences when it comes to folks who use the "I got an offer elsewhere, pay me more and Ill stay". If you give counter offers when that happens (as opposed to across the board bracket increase), you quickly end up with statistically significant discrepancies, and that's not ok.

I remember that in my peak webdev years in 2008, you built webapps by designing the HTML, converting it to templates, filling in data with Django or Rails, and then adding judicious interactivity with JQuery. By 2011 the world had moved on to Angular and SPAs, and you built webapps as a single HTML page and large JS bundle with a bunch of components that you'd fetch data for over AJAX. By 2013 the world moved on to React, and you had all these tools (Gulp, Grunt, Bower, NPM, etc.) to automate packaging and code-reuse. In 2015 people were still using React and more mature versions of these tools, but what changed was the economic reality that you could make lots of money as a webdev, and the industry itself was maturing with demand for new apps dying out.

I'm witnessing this with my team at work too. We have an "old guard" of leadership that joined the team in its peak years in 2018/2019, and learned (and often wrote) the tech stack as it existed then. Now our team is colliding with infrastructure efforts that started elsewhere in the company and were too immature to use last year, but are now starting to bear fruition and get widespread adoption throughout the org. People who were experts in the old way of doing things find that their skills and projects are now largely irrelevant.