---------------------- Forwarded by Vince J Kaminski/HOU/ECT on 03/17/2000 
04:07 PM ---------------------------


Anjam Ahmad
03/17/2000 11:44 AM
To: Ali Lloyd/LON/ECT@ECT, Tom Glover/LON/ECT@ECT, Imtiaz Ahmad/LON/ECT@ECT, 
Christian Hanell/LON/ECT@ECT, Chris Thrall/LON/ECT@ECT
cc: Stinson Gibner/HOU/ECT@ECT, Grant Masson/HOU/ECT@ECT, Dale 
Surbey/LON/ECT@ECT, Bjarne Schieldrop/OSL/ECT@ECT, Vince J 
Kaminski/HOU/ECT@ECT 
Subject: Half-Hourly Option Forward Volatilities

Dear all,

Firstly, apologies for the delay in finalising this study; a considerable 
amount of data analysis was required.  The idea is to aggregate monthly, 
weekly and daily volatilities such that we can price individual options on 
half hours into the future.  We have a reliable methodology for the monthly 
vol curve at the EFA slot level.  The first issue to address was to use a 
"reverse basvol" to estimate the monthly vol curve at the half-hourly level 
of detail.  This required an estimation of half-hourly price return 
correlations.  This resulted in a scaling factor of approximately 1.30 that 
is applied to all EFA vols to convert to half-hourly (the scaling factors are 
slightly different by EFA slot, but not by nearly as much as might be 
expected). 

For the weekly and hourly half hourly vols, a historical analysis was carried 
out on the past few years data; a weighting-scheme was applied to add more 
emphasis to standard deviations calculated on more recent data.  

COMBINING THE VOLATILITIES
The idea is to let the monthly vol curve get us to the start of the month and 
then assume that all half hours are positioned at 2 weeks and 3 days into the 
next month.  The volatility is aggregated using these time-weightings to 
ensure that we don't have a wildly different volatility for the options at 
the start of a month and at the end of a month.  Below is an example for half 
hour #38:

I put the spreadsheet HalfHourlyVolModel.xls into the public directory, 
S:\RESEARCH\ANJAM\HHOPTIONVOLMODEL\. I suggest we use the matrix of forward 
volatilities on the summary page and see what kind of numbers we get.  There 
may need to be some fine-tuning of the time-weightings, but in general, I 
believe we can start utilising these semi-final numbers.

Regards,

Anjam
x35383