The "$9,000 to keep" figure is the cost to keep a disengaged worker on top of their salary. This is the 18% of lost productivity against a $50k average salary. (Which suggests $50k of value creation in a given year dropping to $41k.)

The article doesn't address pay as a way to increase engagement—and thus decrease attrition. Indirectly, the article suggests that increasing pay wouldn't actually have that much effect, with the real benefit coming from managers "who give workers a sense of purpose, inspiration and motivation to perform".

Mathematically, you would need every 2% of average "retention" raise to yield a 1% drop in retention rate to break even, notwithstanding that 18% productivity drop.

Put another way, it's not about employees being commodities. It's about (generally) pay-for-retention programs NOT ONLY failing, but in the worst case negatively affecting those people who are engaged by forcing them to continue to interact with disengaged people who decided to stick around a little longer.