A. D. Smallwood, and S. C. Norrbin.  "Generalized Long Memory Processes,
Failure of Cointegration Tests, and Exchange Rate Dynamics", Journal of
Applied Econometrics, Vol. 21, No. 4, 2006, pp. 409-417.

This study employs daily exchange rate data from March 3, 1980 to
December 31, 1998.  In analyzing the subsample period from March 3, 1980
to February 19, 1985, we use the data employed by Baillie and Bollerslev
(1989,1994, Journal of Finance), which yield $ prices of the foreign
currency.

Alternatively, the exchange rate data for the full sample were
originally taken from the St.Louis Federal Reserve Board (FRED) and
include NOON buying rates of the dollar for the Canadian $, Deutsche
mark, French franc, Italian lire, Japanese yen, Sterling pound, and
Swiss franc. In estimating a GARMA(0,0) model for the full sample (see
footnote 4), we employ the data taken from the St. Louis Fed.  The data
are contained in the following files:

bollerdata.txt
    Date: Numerical value for the date 
    $/C$: US dollar price of the Canadian dollar
    $/FF: US dollar price of the French franc
    $/DM: US dollar price of the Deutsch mark
    $/Lire: US dollar price of the Italian lire
    $/Yen: US dollar price of the Japanese Yen
    $/SF: US dollar price of the Swiss franc
    $/Pd: US dollar price of the British pound

dailERdata.txt
    Date:  Numerical value for the date 
    C$/$:  Canadian $ price of the US dollar
    FF/$:  French franc price of the US dollar
    DM/$:  Deutsch mark price of the US dollar
    Lire/$: Italian lire price of the US dollar
    Yen/$: Japanese Yen price of the US dollar
    SF/$:  Swiss franc price of the US dollar
    Pd/$:  British pound price of the US dollar
    
These two files are ASCII text files in DOS format. They are zipped in the
file sndata.zip. Unix users should use "unzip -a".

NOTES: The estimation of the GARMA(0,0) model was conducted using an
algorithm the authors created in MATLAB.  The algorithms for the Monte
Carlo study using the Engle-Granger (1987) and Johansen (1988)
techniques were also carried out using MATLAB code created by the
authors.  The authors will make this code available on an individual
request basis.