Martin Martens, Paul Kofman, and Ton Vorst, "A Threshold Error Correction
Model for Intraday Futures and Index Returns", Journal of Applied
Econometrics, Vol. 13, No. 3, 1998, pp. 245-263.
Data source
***********
Futures Industry Institute
2001 Pennsylvania Avenue N.W. Suite 600
Washington, D.C. 20006-1807
USA
(tel: (202) 223-1528; fax: (202) 296-3184
Note that the original data can be purchased from the FII. Below follows the
description of the transformed data as used in the paper.
Data construction
*****************
For the months May and November 1993 transaction data were obtained from FII.
These include the S&P 500 index spot calculated every 15 seconds, and the
futures on the S&P 500 index showing the time and the price for each
transaction.
The last price in each minute is used to form the 1-minute price series.
The first 10 minutes of every trading day are removed, and also no overnight
returns are included. For each day there are then 379 returns, unless trading
halted earlier. All these prices/returns are stacked into one vector.
All data files are in DOS format (CR/LF pairs). They are zipped in the
file mkv-data.zip.
Data-files index and futures prices/returns
*******************************************
November 1993 (each datafile contains 1 column, having 7693 observations)
futurnov.asc: Futures prices
spotnov.asc: Spot prices
dfutnov.asc: Futures returns (same length as prices; corresponding minutes)
dspotnov.asc: Spot returns
May 1993 (each datafile contains 1 column, having 7060 observations)
futurmay.asc: Futures prices
spotmay.asc: Spot prices
dfutmay.asc: Futures returns
dspotmay.asc: Spot returns
Data-files cost-of-carry (coc)
******************************
To construct the error correction term as defined in equation (2) in the
paper interest rates and dividends are needed. Dividends are reported in the
S&P 500 Information Bulletin and were kindly provided by Bob Whaley. For the
interest rates we used the daily US discount rate interpolating between the
rates for the various maturities. These data were taken from Datastream.
The resulting daily cost-of-carry term is then adjusted according to the
number of observations per day to match the price series.
November 1993 (each datafile contains 1 column, having 7693 observations)
tfutnov.asc: Theoretical futures prices (index prices adjusted for coc)
ectnov.asc: Error correction term (same length as prices; corresponding
minutes)
May 1993 (each datafile contains 1 column, having 7060 observations)
tfutmay.asc: Theoretical futures prices
ectmay.asc: Error correction term
Raw data
********
The original raw data can be purchased from the above described sources.
Upon purchasing we had to sign the agreement the data would not be given to
third parties. The data provided here allow replication of all results
produced in the paper.