Achim Zeileis and Christian Kleiber, "Validating Multiple Structural Change
Models -- A Case Study", Journal of Applied Econometrics, Vol. 20, No.
5, 2005, pp. 685-690.

INTRODUCTION

This paper is a replication study of the results obtained in

  Jushan Bai and Pierre Perron, "Computation and Analysis of Multiple
  Structural Change Models", Journal of Applied Econometrics, 18, 2003,
  pp. 1-22.
  
who analyze three different models using empirical data and their
implementation of the methodology written in the GAUSS language.

We conducted the replication study based on their GAUSS implementation and
our implementation in the R language. Due to space restrictions, our paper
only describes the replication of the first model in detail; the other two
models are only briefly mentioned.

The full replication study analyzing all three models is available
as a technical report

  Achim Zeileis and Christian Kleiber, "Validating Multiple Structural
  Change Models -- An Extended Case Study", Report 12, Department of
  Statistics and Mathematics, Wirtschaftsuniversitaet Wien, Research Report
  Series. URL http://epub.wu-wien.ac.at/
  
Here, we provide the data sets (in the file zk-data.zip), the software used
(in the file zk-software.zip), and code/output for all three models and for
both GAUSS and R (in the files zk-gauss.zip and zk-r.zip). With the
exception of the tar and zip files in zk-software.zip, all files are ASCII
files in DOS format.


DATA

The data sets for the three models analyzed are provided in space-separated
files:

  1. real.dat
     US ex-post real interest rate from 1961(1) to 1986(3)
     
  2. cpimat.dat
     Data for the AR(1) model for the log-price index p_t in Equation
     (12) (in Bai and Perron's paper). The file contains the first
     differences and the lagged first differences of p_t from 1948 to
     1987.
     
  3. pcmat.dat
     Data for the Phillips curve equation (13). The five columns
     correspond exactly to the variables required for fitting this
     equation: first differences of w_t (log wages), a constant, lagged
     first differences of p_t, first differences of u_t (unemployment
     rate), lags of u_t.
     
The data for 1. is also available in the R package strucchange via
data(RealInt) and the data for 2./3. via data(PhillipsCurve).
More detailed descriptions are also available on the corresponding
manual pages.


SOFTWARE

In the zip file zk-software.zip, there is a directory "GAUSS" with the
GAUSS source code and a directory "R" which provides the R packages
used.

GAUSS

Two versions of the GAUSS code are provided: 1. the original code
brcode.src and 2. our modified code bp-mod.src.
The original brcode.src was taken from the JAE data archive at

  http://qed.econ.queensu.ca/jae/2003-v18.1/bai-perron/

Our modification fixes several problems described in our paper.
By now, Pierre Perron also fixed these problems in brcode.src,
the most current version is available from his Web page at
  
  http://econ.bu.edu/perron/code.html

R

All the results in our paper have been computed using the R system
for statistical computing (http://www.R-project.org/) and in
particular the R package strucchange (see Zeileis et al., 2002) for
testing, dating and monitoring structural changes. Both the R system
and the strucchange package are freely available at no cost under the
terms of the GNU General Public Licence (GPL) from the Comprehensive
R Archive Network at http://CRAN.R-project.org/

The results in the paper have been obtained using strucchange 1.2-7,
sandwich 1.0-0 and zoo 0.9-1 which are also available in this data
archive as source code (*.tar.gz) and as compiled binaries for Windows
(*.zip). Current versions of all packages can always be obtained from
CRAN.

The R version used to obtain the results in the paper was 2.0.1.


CODE/OUTPUT

In the zip files zk-gauss.zip and zk-r.zip, the code and output for the
models described in the Bai and Perron paper are provided. The prefixes are
Table1, Table2, Table2-h7 and Table3 corresponding to the tables in the
original Bai and Perron paper (and the three data sets described above).
Table2-h7 contains the modified analysis of Table 2 employing the parameter
setting h = 7 (instead of h = 8).

The R code and output have suffixes .R and .Rout, respectively. For GAUSS
the suffixes are .prg and .out, respectively.

Additionally, the original Bai and Perron script break.prg along with
its output break.out are provided.