In companies, perception is more important than reality - this is why easily measured variables are treated as more important than difficult-to-measure ones - you can justify decisions by pointing to hard data even if that data is barely related to the problem at hand. Imagine trying to justify giving a developer a large raise (a measurable impact) and justifying it with an argument that retention is probably more profitable than churning (hard to measure). You'll be laughed out the door.

People aren't rational actors in the sense of "they aim for the objectively best action given a goal" - we generally aim to minimise risk, prefer things that are simple, and prefer doing what everybody else is doing. Homo economicus isn't even used in modern economics let alone being anywhere near to reality.

> Imagine trying to justify giving a developer a large raise (a measurable impact) and justifying it with an argument that retention is probably more profitable than churning (hard to measure). You'll be laughed out the door.

It depends how supply and demand curves are moving. What might be true today may not have been true 10 years or 20 or 30 years ago. Labor costs have a big impact on profit, a measurement that all company owners and executives care very much about.

For the past few decades, it may have been true that the cost of attrition was less than the benefits of holding down pay for most others.

That's a post hoc argument - you're presenting a possibility to explain why not minimising attrition was actually rational all along. Its quite well established by now that human decision making is not very rational. Even people who aim for rationality are generally biased towards variables with high measurablility.

The reasoning is that if preventing attrition was so valuable, then it would have shown up via better profits for various businesses over the many previous decades, especially publicly listed ones since their financials are public.

I am sure some businesses got it wrong, and some businesses got it right. And it probably even varies within businesses based on specific roles.

For example, it very well may be rational for a grocery store to let cashiers go if the cost of acquiring and training a new one is low enough that it lets the store compete with other grocery stores since they are operating on razor thin margins. On the other hand, people with specialized tasks in the store such as a butcher or baker or manager might have a better case for keeping them from leaving.

That's the perfect market fallacy, which is also a post hoc justification.

I think right now it is vital to double-digit raise your engineers if needed, because hiring is super difficult right now. You not only stand to lose an employee, you stand to lose the profit of having any employee there for months.

Yes there's a lot of money in retaining talent at any cost. That's what FANG companies are basically. The only problem being these firms are so good at making money they've sucked out a lot of the oxygen.

Most companies aren't as good at using their retained talent to make money.

Especially with remote work breaking down geographical moats at least within countries/regions, it's entirely possible that for most people salary-maximizing, they go with one of the big West Coast employers (or something like an HFT) if they can get in. And most other companies compete to hire everyone else.

Of course, hiring practices are sufficiently random in many cases that it's not really a matter of those firms skimming all the cream (and salary is only one factor that many go by anyway). But it is harder for other companies to salary match the likes of Google and Facebook.

>So if you believe in humans as rational economic actors,

1) They're not.

2) Even if they were, it may be the case they are short-term biased, meaning they're willing to forgo $20 tomorrow for $10 today.