Gael M. Martin, Andrew Reidy, and Jill Wright, "Does the Option Market
Produce Superior Forecasts of Noise-Corrected Volatility Measures?"
Journal of Applied Econometrics, Vol. 24, No. 1, 2009, pp. 77-104.
There are two zip files. The file AVM-xls.zip contains 8 Excel files, and
the file AVM-csv.zip contains 8 ASCII text files in CSV format. The latter
are DOS-formatted, so Unix users should use "unzip -a". Each zip file
contains eight files:
AVM-xls.zip:
GE-avm-1996-2001.xls
GE-avm-2001-2006.xls
IBM-avm-1996-2001.xls
IBM-avm-2001-2006.xls
msft-avm-1996-2001.xls
msft-avm-2001-2006.xls
sp500-avm-1996-2001.xls
sp500-avm-2001-2006.xls
AVM-csv.zip:
GE-avm-1996-2001.csv
GE-avm-2001-2006.csv
IBM-avm-1996-2001.csv
IBM-avm-2001-2006.csv
msft-avm-1996-2001.csv
msft-avm-2001-2006.csv
sp500-avm-1996-2001.csv
sp500-avm-2001-2006.csv
The corresponding .xls and .cls files contain the same data. There are two
sets of files:
Set 1: Daily values of annualized realized variance measures for the
S&P500 index, International Business Machines (IBM), Microsoft (MSFT)
and General Electric (GE) over the period 1/7/1996 to 29/08/2001.
Each of the .xls and .cls files has 9 columns:
Col. 1 -- Date (day t)
Col. 2 -- RV(5) on day t
Col. 3 -- RVA(5) on day t
Col. 4 -- TSRV on day t
Col. 5 -- TSRV2 on day t
Col. 6 -- RKERN on day t
Col. 7 -- OSRV on day t
Col. 8 -- ALTM on day t
Col. 9 -- BV on day t
Notes:
1. The explanation of all acronyms and the equations in the paper to
which they relate are given in Section 3.1 of the paper. All details of
the manner in which the alternative variance measures (columns 2 to 9)
are constructed are given at appropriate points in the paper, most
notably in Sections 2, 3.1 and 4.1. All realized variance measures have
been transformed to represent 24 hour variance measures, as described in
Sections 3.1 and 4.1.
2. The first year of observations (1/7/1996 to 30/06/1997) was used to
set pre-sample values in the estimation of all long-memory models
3. Rolling samples of size R =3D 1000 (days) were used to produce one- and
22-step ahead forecasts, as explained in Section 4.1, with the first
forecast being for 30/08/2001.
Set 2: Daily values of annualized realized variance measures for the
S&P500 index, International Business Machines (IBM), Microsoft (MSFT)
and General Electric (GE) over the period 30/08/2001 to 31/05/2006, plus
the prior day option-implied forecasts.
Each of the .xls and .cls files has 14 (or 15) columns:
Col. 1 -- Date (day t)
Col. 2 -- RV(5) on day t
Col. 3 -- RVA(5) on day t
Col. 4 -- TSRV on day t
Col. 5 -- TSRV2 on day t
Col. 6 -- RKERN on day t
Col. 7 -- OSRV on day t
Col. 8 -- ALTM on day t
Col. 9 -- BV on day t
Col. 10 -- ATM on day t-1
Col. 11 -- MF on day t-1
Col. 12 -- MF(1.5) on day t-1
Col. 13 -- MF(2) on day t-1
Col. 14 -- MF(2.5) on day t-1
Col. 15 -- VIX (for S&P500 data file only) on day t-1
Notes:
1. The explanation of all acronyms used for column headings 10 to 15 are
given in Sections 3.1, 4.1 and 4.2.4 of the paper. All details of the
manner in which the alternative option-implied variance measures (in
columns 10 to 15) are constructed are given at appropriate points in the
paper, most notably in Sections 3.2.3, 4.1 and 4.2.4.
2. This is the sample used for forecast evaluation (via the SPA test):
3. The option-implied forecasts (Section 3.2.3) of any realized variance
measure on day t of this sample period are constructed from option
market data for day t-1.
4. The "direct" returns-based forecasts (Section 3.2.2) of any realized
variance measure on day t of this sample period are constructed from
time series models fitted to the realized variance measures over the
1000 days up to and including day t-1.
5. The "indirect" returns-based forecasts (Section 3.2.1) of any
realized variance measure on day t of this sample period are constructed
from time series models fitted to the daily closing returns over the
1000 days up to and including day t-1.
All equity data were supplied by the Securities Industries Research
Centre of Asia Pacific (SIRCA) on behalf of Reuters, with the raw data
then cleaned using the methods of Brownlees and Gallo (2005). The VIX
data were extracted from the CBOE website (www.cboe.com). All ATM, BS and
MF calculations are based on the implied volatility surface data
provided by IVOLATILTY (www.ivolatility.com). IVOLATILITY and SIRCA have
give permission for the publication of this data. The VIX data are on
public record.
Gael M. Martin
Monash University, Clayton
Department of Econometrics and Business Statistics
Building 11E, Monash University, Victoria 3800, Australia
gael.martin [AT] buseco.monash.edu.au