Christian Conrad and Onno Kleen, "Two Are Better Than One: Volatility
Forecasting Using Multiplicative Component GARCH-MIDAS Models",
Journal of Applied Econometrics, Vol. 35, No. 1, 2020, pp. 19-45.

All our data and programs are zipped in the file
Conrad_Kleen_2019_Data_Archive.zip. The statistical programming
language we used is R.

The data used in the article (sources and details) are described in
the Section 4 of the main paper and in Appendix F in the online
appendix. 

Data:
-----

Our data set can be found in the subfolder 'data'. Our code uses the
binary rds-versions, but all files are additionally stored as csv files
for the convenience of users that don't use R.

There are three different types of data sets:

1) df_recent_releases.rds/csv: This a consolidated data set that does
   not contain real-time macroeconomic information.
   
2) df_'macroeconomic variable'.rds/.csv.gz: These are the data files
   that contain the real-time data set for each macro variable. csv
   files are compressed in order to save space.

3) df_'macroeconomic variable_release_dates.rds/csv': Contains the
   release dates of each macroeconomic variable.

Empirical analysis:
-------------------

All our results and most of the corresponding tables and figures of
the empirical section are automatically reproduced by

1) Calling master.R first. Please make sure that the parallel
   toolboxes in R are working fine. If possible, increase the number
   of cores that can be used to pursue the real-time estimation loop.
   The master script builds the folder structure necessary.
   
2) For reproducing the figures, there has to be a working version of
   LaTeX on the computer running.

3) Please run the scripts stored in the subfolder 'scripts'
   consecutively as suggested by their naming scheme. Some scripts
   rely on output generated by a previous one.
   
4) All tables are stored in tex-files that can be included in a main
   tex file as tabular environments.  They can't be compiled on their
   one. LaTeX packages necessary are siunitx and booktabs.
   
NOTE: The real-time estimation strategy will take long if it isn't run
on a computing cluster.

Onno Kleen
onno.kleen [AT] awi.uni-heidelberg.de