Subject: re : risk metrics
i completely agree with you .
nevertheless , i am sorry if i did not make myself clear when we talked about
this last tuesday : i am not advocating penalizing business units because of
their lweek rovars or sharpes . i would certainly not set traders ' bonuses
based on these numbers ( nor do i think anyone at enron would consider doing
that ) .
enron focuses on long term profit and risk . however , the long term is made
up of many short terms , and catastrophes can occur if a company ' s management
does not pay close attention to the latter . indeed , enron management is very
interested in short term profitability ( they want to be notified of very
large 1 and 5 day losses ) and short term risk ( we report daily var limit
violations ) . in this spirit , i am simply offering 1 and 4 week rovars and
sharpes as something interesting and useful that we can now produce from the
dpr database , and i will let upper management use or ignore them as they see
fit .
regards ,
eugenio
naveen andrews @ enron
09 / 08 / 2000 11 : 13 am
to : eugenio perez / hou / ect @ ect
cc : ted murphy / hou / ect @ ect
subject : risk metrics
eugenio ,
in regard to your recent e - mail concerning rovar , sharpe and other
risk metrics , and our meeting on tuesday , it is imperative to understand that
( 1 ) while p & l is a number one can aggregate on a daily basis , var numbers on
1 - week or 4 - week basis , and consequently rovar numbers , are not meaningful
operationally ( ie , trading activity and senior management decision - making )
and statistically . rovar and sharpe were designed to measure business unit
performance over extended time periods ( preferably a year ) . specifically ,
you cannot penalize , or make reasonable business decisions about a unit ,
because of their 1 - week rovar or sharpe .
( 2 ) significant trading activity at ene is composed of long - term strategic
trades ( seasonal plays with durations of over 6 months , perhaps ) , wherein
desk heads and senior management have an intuitive feeling for their rovar
( over 6 months ) . a 1 - week or 4 - week " interim " number in this case is not
useful . traders might purposefully want to make a certain 1 - week or 4 - week
rovar low in their trading activity if they believe their 6 - month number can
be high . in such a case , a 1 - week number might send dangerous messages to
senior management .
( 3 ) conceptually , p & l and var , in isolation , are numbers which are sound and
rigourously calculated . the ratio , however , one has to be careful about .
the numerator ( cumulative p & l over a certain period ) can be aggregated by
simple arithmetic summation , no problems . however , the denominator ( average
daily var over a certain period ) relies on the underlying assumption that
each point in your time series is independent and correlationless . to get a
sufficient statistic for the denominator , one would have to take a large
sample in your series ( 6 months or more ) , for a daily average var number to
be meaningful , statistically .
( 4 ) the rac group has instituted similar metrics which have been in usage
for over two years . the numbers are shown to the b . o . d with the express
intent of comaring business unit performance on a 6 - month or yearly tenor .
please contact matthew adams ( rac ) or myself to make sure that we are on the
same footing regarding the actual calculation of such metrics . also , in the
future , just to be consistent and on - the - same - page , please contact me or mr .
adams if any risk - related numbers are to be showcased in a public forum .
( 5 ) finally , as you can probably corroborate , senior management might not
be fully versed on the nuances of var , let alone 1 - week var ratios .
therefore , one has to be wary of the misuse of ratios and risk metrics .
i liked your presentation of your database on tuesday . i think it is a nice
step in aggregating and dissecting p & l from disparate sources .
best regards
naveen