It all makes "business" sense for those who see employees as "commodities", i.e. all kinda equivalent and hence easilly replaceable with nothing lost when they're switched.
It's basically the MBA thinking of employees as just another "raw material" or "supplier".
The reality, more so in complex domains, is that employees have an adaptation and learning period when they arrive (unlike engineered devices, companies aren't standardized machines using standardized parts, so you a new "part" won't just seamlessly fit and start delivering full performance) and often never written institutional knowledge that goes with them when they leave.
However as those things are not easilly quantified and measurable, MBA types - being unable to add it to their spreadsheets - will simply ignore them rather than trying to balance such costs against salary costs: giving a decent salary increase (a guaranteed cost) will always look like a worse option in an accounting spreadsheet if its only counter is a sub-100% possibility that they might lose that employee (and, remember, since they don't count adaption and loss of institutional knowledge costs, that's listed there as costing nothing) and replace it with somebody else who might even be possible to get with a less "decent" salary (so, more than the current employees but less that a fair salary for the current employee).
Such approach works well if all companies are doing it and the probability that people will leave if they don't get a decent salary is low enough (which it probably is, since the majority of human beings favour stability over change).