Subject: short term private firm model : static historical
snapshot + performance data for model development
mike , scott , eric ,
after brainstorming and discussing further on data here , we think that our
final specifications for modelling data requirement need to be as follows :
we need bankrupt , default and nondefault ( which covers nonbankrupt ) accounts
with 4 quarterly observation snapshots and 12 months performance following
the latest snapshot . monthly performance indicator need to be available for
the entire 12 months performance period . we will need all the bankrupt and
default accounts and comparable sample of good accounts with weights .
for the purpose of model validation , we will need data with above specs
covering 16 months of performance . this means that we will need rolling ( 4
quarterly snapshots + 12 months performance ) data for 4 monthly shifts :
input snapshots performance
1999 march end , 1999 june end , 1999 september end , 1999 december end 12
month end performance for jan 2000 through dec 2000
1999 feb end , 1999 may end , 1999 august end , 1999 november end 12 month
end performance for dec 1999 through nov 2000
1999 jan end , 1999 apr end , 1999 july end , 1999 october end 12 month end
performance for nov 1999 through oct 2000
1998 december end , 1999 mar end , 1999 june end , 1999 september end 12
month end performance for oct 1999 through sep 2000
we will need bankruptcy chapterwise indicator , if available during the
performance period . our definition of default is either bankruptcy or 90 +
days delinquency on any trade credit .
we have also discussed the cost aspect and we think considering the big
picture , it makes sense to spend the extra amount to get the right data for
analysis .
please let me know your thoughts . this will require d & b to give us a
modified quote and we could possibly move forward quickly .
regards ,
amitava