I think you view (or at least present here) the programming job as pure production measured in KLOCs or something similar, and this is wrong. It is a part of the job, but not the main part. 1 KLOC written by a good programmer may be worth literally orders of magnitude more than 10 KLOC written by 10 poor programmers. The reason is not syntax errors - those are mostly irrelevant as the tools get rid of them pretty quickly. The relevant part is what tools can't do - conceptual build of the code, design, robustness and ability to accommodate change. These terms are very hard to define, especially with rigor that would be convincing to anybody outside of the industry, so there's a lot of hand-waving and "you know what I'm talking about" when we talk about it, but in the essence poor code is poor and great code is great not because the former has tons of obvious errors and the latter does not. It is because you'd spend orders of magnitude more money, time and other efforts supporting and improving poor code than great code. That's where better programmers are providing the edge, not in the KLOC output. If you need KLOC, you don't need great, you don't even need good, you just need "can tie his own shoes" one. But if you need a system that you wouldn't have to throw out and rewrite in 3 years because it became unmaintainable and every fix immediately creates an explosion of new problems - then you need great. That's where 9 women/1 months comes into play.
The problem here is of course as I noted this is very hard to manage and evaluate except way post-factum (and even then there are little of robust tools and methods to do it with any rigor) so it's hard to know for sure. But in many cases, "you know it when you see it".