Hans Dewachter, Marco Lyrio, and Konstantijn Maes, "A joint model for the
term structure of interest rates and the macroeconomy", Journal of Applied
Econometrics, Vol. 21, No. 4, 2006, pp. 439-462.

The two files described below are zipped in the file dlm-files.zip. They
are both ASCII files in DOS format. Unix users should use "unzip -a".


1. DATA (data.txt)

The data set is provided in the file data.txt and consists of 164
quarterly observations for the period 1958:Q1 to 1998:Q4. We only make
use of 140 observations for the period 1964:Q1 to 1998:Q4, due to the
unreliability of long-term bond data for the period before 1964.

The structure of the data is the following:

--- yields by maturity ----
3m  6m  1yr  2yr  5yr  10yr   output gap   inflation

The term structure data were provided by Gregory Duffee and used in the
paper "Term premia and interest rate forecasts in affine models"
(Journal of Finance 57, 2002, 405-443).

The output and inflation series are from the International Financial
Statistics (IFS) database provided by the International Monetary Fund (IMF).

A proxy for the output gap is obtained by using a Hodrick-Prescott (HP)
filter on the GDP series using a standard "lambda" in the filtering
procedure equal to 1600.

Inflation is constructed by taking the yearly percentage change in the CPI
index.


2. PROGRAM (program.prg)

The file program.prg contains the Gauss program used in the estimation of
the macro model. We used version Gauss 3.2 to run the program. The library
CML (constrained maximum likelihood) is needed to run the program.

The program is prepared to run in the directory c:\jae\

Both files "program.prg" and "data.txt" should be copied in this directory.
In order to change this directory, one should redefine the variable
"filedir" at the beginning of the program.

When the program is run, the file "bestever.txt" is created in the same
directory. This file contains:

- the settings for the estimated model;

- a vector with the input variables relative to the best results until that
  moment (can be copied into the file "bestbnew.txt");

- the same results but in the original vectors and matrices (can be copied
  inside the program);

- a table with the improvement for each of the estimated variables; and

- the set of results as they appear on the screen.

The file "bestbnew.txt" is also created when the program is run. It
contains the same vector with the input variables relative to the best
results until that moment as in the file "bestever.txt".

Both files "bestever.txt" and "bestbnew.txt" are updated at the number of
iterations specified by the variable "printCML".

The input variables can be read from two possible sources:

- From the values defined inside the program. The variable "filebestb"
  should be set to zero. This is the default of the program;

- from the file "bestbnew.txt". The variable "filebestb" should be set to
  one.

The optimal parameter estimates are included in the program.