Accounting and law firms are looking for people with the stamina and single-minded goal focus to get document-creation tasks done, one after another, to a high degree of "skill." Getting an A on the test and preparing an SEC-compliant bond prospectus are close cousins of each other.
In fields where real-world mastery doesn't have much to do with the orderly routines of getting a 97 on the midterm, employers are less likely to hunt for people who got 97s on the midterm. (John Steinbeck stumbled around Stanford for six years and never graduated. But he got more accomplished as a writer than a whole decade worth of creative-writing majors with high GPAs.)
Are they really, or is just this elaborate illusion supported by most decision makers to give some rationalization to their own selection processes? When you have Harvard being sued for discriminating against Asians, how can we keep a straight face and tell the kids that they need to study hard because the grades matter?
I think this is part of what PG wants to defend in the essay: why do we actually care about grades at all?
Because jobs are at a premium and evaluating prospective hires is hard, so employers look for anything they could use to filter and sort the stream of candidates. Education - both the what and the where is somewhat correlated, so that's what's being used.
I think I spent too much on this discussion already, but the point I am trying to make is that standard tests are bad both as a filter for selection and as predictor of performance.
While not necessarily agreeing, I do understand if someone says "I will hire someone good from a top-tier school vs a top student from a not-so-good university". The "what" and "where" are definitely useful data points to filter out candidates.
What I don't understand is any employer that says they will hire based on SAT, GPA or any variation of an IQ test. It seems to be so unreliable measures to the point of actually hindering the process.
We could give more autonomy to individual schools or teachers in terms of designing tests, but then we'd be facing two problems. One, the grades given by different schools/teachers would not be directly comparable. An A in one school could be awarded only for exceptional circumstances, and in another just for above-local-average. Two, once you start using those grades to judge people in a competitive situation like the job market, you'll have schools selling grades, teachers inflating grades to help their pupils compete against those from schools that sell grades, etc. and the whole system will collapse to uselessness.
Standardized testing mitigates these problems by forcing everyone to grade by the same criteria, but in order to do that, the grading must be done on things that are extremely easy to measure - which removes both individual judgement of the teachers and any relation between testing and understanding of the material.
The more I think about it, the more I feel it all boils down to scale. The grading system needs to be scaled up with the size of the market people compete on, otherwise it'll be either trivially gamed, or deeply unfair to unlucky people.