It's all relative, though. If you increase salaries for retention, it has to come out of some other pot. Maybe that isn't your new hire pool, but you're going to see a negative metric of some kind somewhere. Now don't get me wrong, I still think it's better in the long term. But my point is that you'll get some kind of negative signal by pulling back funds elsewhere, and a firm at that point has to have the discipline not to freak out and stay on course.
If you're correct, and it costs more money to keep people, then we are wrong: it's not cheaper to go for stability.
Maybe it's like in the prisoner dilemma: you can't go lower than the market peak because no-one will come to you, even if you offer more on average than the industry average. This way, we reward the job hoppers with the peak salaries. Peak salaries means over time the average goes up too. I guess it's how free markets operate.
It depends on how retention interacts with other variables. I would argue that you can make exponential gains in things like product quality over the long term by retaining a critical mass of experienced personnel, which translates to higher earnings etc. But yes if you view personnel costs as a line item in isolation and on a quarterly basis, you might not see those effects.
