Thanks to @asg
who discovered this text, I am reproducing it below:-
Scaling in UPSC Mains Explained
The head examiner
is also the paper setter
If the number of papers to be examined are less than 350
, the head examiner checks all the copies but if the number exceeds this limit then for each additional 300 copies an additional examiner
is appointed. The head examiner convenes a meeting of all the examiners
in the UPSC premises within seven days
of the conduction of an optional subject. All of them are asked to prepare model answers for all parts and sub parts of the question paper. These questions and answers are then thoroughly discussed. At the end of the discussion the head examiner with the consensus of the additional examiners issues final guidelines.
After this the examiners examine all the copies within a specific time frame
.Once this is over, all the copies are arranged in vertical stacks (examiner-wise) with the head examiner. The head examiner then conducts a sample survey
In the sample survey he chooses 20 copies
from each stack (examiner). These 20 copies are - 5 which scored the highest
in the stack; 5 which scored the lowest
in the stack and 10 selected randomly
. The head examiner then conducts upward or downward moderation
of the marking of these copies, if the need be. If the discrepancy or deviation in any of the stack(s) is found erratic, he moderates the entire stack
Once the subject scrutiny is over the UPSC resorts to a statistical process known as 'normative linear regression'
and brings all the subjects on a common plane. In the case of GS there is no scaling as such, although the rest of the process, before that, remains the same.
In the case of essay the process is entirely different.
There are no head examiners. There are a score of examiners appointed 2 per 300 copies. Each copy is evaluated by two examiners and the mean score of the two scores is taken into consideration. the marks awarded to a script by the first examiner is kept confidential from the second examiner.