Qianqiu Liu, "On Portfolio Optimization: How and When Do We Benefit from
High-Frequency Data", Journal of Applied Econometrics, Vol. 24, No. 4, 
2009, pp. 560-582.

All the data used in this paper are confidential.

The high-frequency data used in this paper are from the Trade and Quotation
(TAQ) and Chicago Mercantile Exchange (CME) databases. The TAQ data files
contain continuously recorded information on the trades and quotations for
the individual stocks listed on the New York Stock Exchange (NYSE), American
Stock Exchange (AMEX), and the National Association of Security Dealers
Automated Quotation system (NASDAQ). The database is published monthly and
has been available since January 1993. The sample period examined in this
study is from January 2, 1993 to June 30, 2000, a total of 1894 trading
days. The sample consists of all of the quotes for the 30 DJIA firms as of
the reconfiguration of the DJIA index on November 1, 2000.

Five-minute returns are constructed from the sample. The daily transaction
record extends from 9:30 EST until 16:00 EST. Each quote consists of a bid
and an ask price along with a "time stamp" to the nearest second. To avoid
potential biases and reporting errors in the price quote data, which may
occur at the opening of the exchange, the first 30-minute window of data is
removed, and this investigation includes only price quote data after 10:00
EST. Hence, the intraday sample covers the records from 10:00 EST to 16:00
EST, resulting in a total of 72 five-minute returns for each trading day.
The five-minute return series for the 30 stocks in DJIA are constructed from
the difference between the average of the log bid and the log ask prices
that are recorded at or immediately before the corresponding five-minute
delineation. Overnight returns are computed accordingly as the difference
between the mid-points of log quotes at 10:00 EST and the previous 16:00
EST, adjusted for dividends and splits.

Intraday observations on the S&P 500 cash index are obtained from CME. The
period covered by the data is the same as for the 30 DJIA stocks, from 9:00
CST to 15:00 CST each trading day from January 2, 1993 to June 30, 2000.

The five-minute return and overnight series of the S&P 500 index are
similarly constructed, except that they are the logarithmic differences
between the cash index levels at each five-minute and overnight interval.
The daily returns of the 30 DJIA stocks and the S&P 500 index of the same
period can be obtained from the Center for Research in Security Prices
(CRSP) daily data files. Both are adjusted for dividends and splits. None of
the databases allow open access.

The variables in the sample include month, date, year, hour, minute, second,
bid, and ask. The file "msft1.m" is a Matlab program to generate the bid and
ask prices for stock "msft" in each five-minute interval.

Please address any questions to:

  Qianqiu Liu
  Shidler College of Business
  University of Hawaii
  2404 Maile Way, E-602f
  Honolulu, HI 96822
  E-Mail: qianqiu [AT] hawaii.edu