If you're firing "the worst", that means at some point, you hired them. Not to mention "the best" is a very vague term. How do you quantify quality? What pool of people are you selecting "the best" from? How many of them are "the best"? The first two? The first ten?
Firing gets rid of the worst at playing the internal political game.
Fixed it for you.
Something I've seen at at least a couple of companies where it resulted in killing the company, and one of those was the storied Lucent (forced to sell out to Alcatel, which itself didn't fare well, evidently most of it was sold to Nokia in January this year).
The managers even made a point of telling us how their firing a particular key guy showed no one was safe. Well, the same was true of the project, which was de jure and de facto necessary for Lucent's continued existence as an independent entity, it was a media gateway, which along with a media gateway controller such as their very well received Softswitch replaces a unitary switch like their 5ESS.
(It was a very interesting project, BTW, we were working with 5ESS types in Cincinnati, and like that storied system it was a true 5 9's system, no faking it for scheduled maintenance, people expect their telecom systems to be available 24x7 absent acts of God.)
Healthy companies at scale and size equilibrium should be doing both at roughly the same rate, averaged over time.