Subject: re : uk power / gas
vince , just fyi :
oliver ( risk control in london ) was asking if it is appropriate to use a set
of factors corresponding to some commodity for another commodity .
this " mapping " we do quite frequently in our var system when forward prices
for a commodity are not correlated ( because of poor price history ) .
i think using this kind of mappings is ok because we can not trust these
correlations based on illiquid price information .
if we believed the matrix with low correlations represents what goes on in
the market and we simply can not reconstruct it with our 7 factors - then
we might use more factors then 7 or the matrix itself .
tanya
tanya tamarchenko
12 / 15 / 2000 02 : 36 pm
to : oliver gaylard / lon / ect @ ect
cc : david port / market risk / corp / enron @ enron , rudi zipter / hou / ect @ ect , wenyao
jia / hou / ect @ ect , kirstee hewitt / lon / ect @ ect , debbie r brackett / hou / ect @ ect ,
naveen andrews / corp / enron @ enron
subject : re : uk power / gas
oliver ,
i completely agree with you : validating var inputs like positions , prices and
volatilities which are used by risktrac is the first thing to do .
you also rise a valid question : if the factor loadings for some commodity do
not make sense should we use the factor loadings for another commodity ?
the factor loadings " do not make sense " when the correlations across the term
structure are not high . so if we believe that the forward prices for this
commodity are market prices and statistical analysis on these prices produces
the correlations which we trust , then it would be proper to use this
correlation matrix in var engine , not the factors . using factors is simply
the way to speed up calculations for highly correlated prices without
sacrificing the accuracy .
tanya .
oliver gaylard
12 / 15 / 2000 07 : 11 am
to : naveen andrews / corp / enron @ enron
cc : david port / market risk / corp / enron @ enron , rudi zipter / hou / ect @ ect , wenyao
jia / hou / ect @ ect , kirstee hewitt / lon / ect @ ect , tanya tamarchenko / hou / ect @ ect ,
ganapathy ramesh / hou / ect @ ect , debbie r brackett / hou / ect @ ect
subject : re : uk power / gas
naveen
regarding the calculation of uk vars in risk trac i agree that we should be
using this calculation engine for all commodity vars . however we should not
focus solely on the uk but ensure that we use risk trac for continental power
and gas , uk power and gas , nordic power .
to use risk trac i think the following need to be resolved first , to
implement it " right the first time " , as i think it is incorrect to consider
the risk trac numbers " as the most accurate " since it depends on the validity
of these items :
positions ( delta and gamma ) and curve mapping - these need to include all
positions including those outside the main risk systems
positions , curves and mapping should now be no problem given the feeds ,
apart from continental power , have been uat ' d and we have the spreadsheet
feeds up and running .
price and vol curves - as used by the risk systems
as above
inter commodity correlation - prompt month
correlation should be easy to calculate given an accurate and complete data
set ( however the incomplete historic data for europe in risk trac , prior to
the formation of the task force , would mean a full data set needs to be
obtained and used ) . term structure of correlation would be good but i
understand this is difficult to use in the calculation .
factor loadings
i think factor loadings should be calculated , on the same data sets used
for inter commodity correlation , for all commodities . if this analysis does
not appear to work i am not sure that using factor loadings for other markets
is adequate . do we need to consider an alternative approach to calculating
var for these markets ?
to ensure this moves forward i think a list of the mile stones ,
responsibilities and time lines needs to be drawn up otherwise i fear the
process of moving across to risk trac from the spreadsheet var will
experience some slippage .
i will call today to start the process off .
rgds
oliver
naveen andrews @ enron
06 / 12 / 2000 21 : 50
to : oliver gaylard / lon / ect @ ect
cc : david port / market risk / corp / enron @ enron , rudi zipter / hou / ect @ ect , wenyao
jia / hou / ect @ ect , kirstee hewitt / lon / ect @ ect , tanya tamarchenko / hou / ect @ ect ,
ganapathy ramesh / hou / ect @ ect , debbie r brackett / hou / ect @ ect
subject : uk power / gas
oliver ,
i had a couple of issues pertaining to uk power / uk gas . first ,
just a few notes as it pertains to risk trac uk - implementation :
( 1 ) in risktrac , currently all the gas curves are mapped to nbp .
( 2 ) all the power curves are mapped to r 4 ( cinergy ) .
factor loading analysis lends itself only to nbp and the norewgian curves ( for
reasons of liquidity , etc ) . we have decided that nbp and cinergy are the
best curves available at this time .
the spreadsheet which is utilized in the uk also has curves mapped to a
2 - 3 - year old set of factors derived from us nat gas , which is clearly not
optimal . hence
( 1 ) we should be using risktrac numbers for var , as it is the " most
accurate " , both in terms of ene - wide model usage and in terms of the most
recent updated data . ever since uk power came on board in risktrac 3 weeks
ago , the uk power number has been consistently 7 - 8 mm above the numbers in
your spreadsheet . this difference is to be ecpected , given the different
inputs .
( 2 ) the consistent numbers do not point to a data error in any obvious way ,
however , if you believe that positions are not captured correctly , please let
our it staff know that .
( 3 ) checks which ramesh has done indicate that positions are tying in .
however , as you know , there could be disconnects with enpower , etc .
in any event , it would be ideal and optimal to have all the simulations run
out of risktrac for reasons of aggregation and analysis .
your help is appreciated .
regards
naveen