Andrea Carriero, George Kapetanios, and Massimiliano Marcellino, 
"Forecasting Large Datasets with Bayesian Reduced Rank Multivariate 
Models", Journal of Applied Econometrics, Vol. 26, No. 5, 2011, pp. 
735-761.

The raw dataset consists of 132 monthly US economic time series observed 
from 1959:1 through 2003:12. All series are from the Global Insights 
Basic Economics Database and the Conference Board's Indicators Database. 
The data set is the one used in

Stock JH, Watson MW. 2006. "Forecasting with many predictors", in 
Handbook of Economic Forecasting, Elliott, G., Granger, C.W.J., 
Timmermann, A. (eds.), Elsevier.

Not all the 132 time series are used in our empirical application.  The 
used series are listed in Table 1 in our paper.  Table 1 in the paper 
lists the series code (which is the series label used in the source 
database) with a brief data description.

All data are in the file swdata.txt, an ASCII file in DOS format.  The 
file has 541 lines and 133 columns.  The first line contains the code 
(i.e. a short name for the series). In the first column there are the 
dates. All the remaining columns contain time series.

The file swdata.txt is zipped in swdata.zip. Unix/Linux users should use 
"unzip -a".