[position in research by date] [position in research by subject]

"Shock Effects on Stocks, Bonds, and Exchange Rates," Journal of International Money and Finance, 2003, 307-341.
pdf file (1,061KB).

Abstract

Tick data and newswire searches are used to find events that led to large and rapid changes in a stock future, a bond future, and three exchange rate futures. Knowledge of these events may be useful in future work. They have the advantages that they are truly surprises and that the sign of their effect on each financial market is known. The events are used in this study to analyze the effects of three types of events (monetary, price, and real) on the five variables.

Comments

The data that were used in this paper are available for downloading. The futures contracts are: S&P 500 (SP), 30 year U.S. Treasury bond (US), deutsche mark (DM), yen (JY), British pound (BP), and euro (EC). In addition, data were collected but not used for the Nikkei 225 contract (NK) and the U.S. T-bill contract (TB). The collection of the data is discussed in Section 2 of the paper. See in particular Table 1 and footnote 2. For downloading, there is one dataset for each contract. For example, the dataset for SP is SP.MIN, which for downloading is in a zipped file named SP.ZIP. There is one line per each minute in which at least one trade took place. For example, the ninth line of SP.MIN is:
 1982  4 21 10 11     11615  -0.043029  -0.086022   1
 
The first number is the year, the second the month, the third the day, the fourth the hour, the fifth the minute, the sixth the price, the seventh the one-minute percent change, the eighth the five-minute percent change, and the last the number of trades in that minute:
 year month day hour minute price one-min-change five-min-change volume

If there were more than one trade in the minute, the price is the last price of the minute. If the one-minute percent change could not be computed because there were no trades in the previous minute, the "missing" observation is denoted 99.000000. Likewise, if the five-minute percent change could not be computed, the missing observation is denoted 99.000000. The times are Central time, not Eastern time as used in the paper. The total number of lines in each file is the total number of available price levels. These are listed in Table 1 and footnote 2. The FORTRAN format of the line is (1X,I4,4(1X,I2),1X,I10,2F11.6,I5). The size of the .ZIP datasets for downloading are:

SP.ZIP  21.8 MB
US.ZIP  11.5 MB
DM.ZIP  12.4 MB
EC.ZIP   0.9 MB
JY.ZIP  14.3 MB
BP.ZIP  10.8 MB
NK.ZIP   1.3 MB
TB.ZIP   1.8 MB
The size of the .MIN datasets that have been zipped are:
SP.MIN  125.4 MB
US.MIN   84.0 MB
DM.MIN   82.9 MB
EC.MIN    5.6 MB
JY.MIN   91.8 MB
BP.MIN   69.2 MB
NK.MIN    9.6 MB
TB.MIN   15.4 MB
These datasets are large, and I have found it easiest to work with them using FORTRAN. They are probably too large for most package programs.

To download a dataset, click it below:

  1. SP.ZIP
  2. US.ZIP
  3. DM.ZIP
  4. EC.ZIP
  5. JY.ZIP
  6. BP.ZIP
  7. NK.ZIP
  8. TB.ZIP