Joscha Beckmann, Gary Koop, Dimitris Korobilis, and Rainer Schüssler,
"Exchange Rate Predictability and Dynamic Bayesian Learning", Journal
of Applied Econometrics, Vol. 35, No. 4, 2020, pp. 410-421.

All files are ASCII files in DOS format. They are zipped in the file
bkks-files.zip. Unix/Linux users should use "unzip -a". Note, however,
that this may destroy the .mat files in CODE/Data_Store/.

The online appendix BKKS_appendix.pdf includes many additional 
theoretical and empirical results, data descriptions, and robustness
checks. 

bkks-files.zip includes data and MATLAB programs.

DATA

The data used in this paper are described in Section 3 in the article
and in the Data Appendix (Section 2 of the online appendix
BKKS_appendix.pdf). If you use these data, please refer to the
original data providers.  See Section 2 of the online appendix
BKKS_appendix.pdf for information about the original data sources.  

Data are in the following CSV (MS-DOS) files:

Variable(s)                                  File name

EXCHANGE RATES                               exrate.csv
UIP                                          uip.csv
INT_DIFF                                     idiff.csv
STOCK_GROWTH                                 stockg.csv
OIL                                          oil.csv
INTEREST RATES                               int.csv
FXVOL                                        fxvol.csv
VIX                                          vix.csv

The data sample ranges from 1973:01 to 2016:12 for the exchange rates,
1986:06 to 2016:12 for VIX, 1993:01 to 2016:12 for FXVOL  
and from 1986:01 to 2016:12 for the reamining variables.

MATLAB CODES

With main.m the econometric model can be generated and results evaluated.

The folder "functions" contains routines that are necessary to execute
main.m:

construct_model_index.m  Construct matrix with all possible model specifications
mlag2.m                  Generate lagged Y matrix
create_RHS.m             Create RHS of VAR model
create_RHS_xx_2.m        Add asset-specific predictors to the RHS
create RHS_xxx.m         Add non asset-specific predictors to the RHS
nearestSPD.m             Find nearest symmetric positive definite matrix  
tvpvarprior.m            Set up the prior 
Minnesota_prior_flexible.m Construct Minnesota prior mean and covariance matrix

Please address any questions to:
raineralexanderschuessler [at] gmail.com