Fixed-Length Binary CRSP File Conversion and Support Code

CRSP ships its stock returns and master/returns files in a 
limited number of file formats. Many subscribers receive the
character format, which takes up a lot of disk space (more
than 2 GB for the 1995 DA product) and is slow to process.
CRSP provides subroutines that can be used to convert the 
character format to a binary format that is much more compact.
Unfortunately, SAS cannot correctly read the binary format on
most Unix and Windows systems and some VMS systems. The basic
problem is that SAS misinterprets some binary data as end-of-
record characters.

Arnold R. Cowan, Associate Professor of Finance at Iowa State
University and President of Cowan Research, has designed an
alternative binary CRSP file format. It is fixed length for
compatibility with SAS, but as a flat-file binary format it
preserves the advantages of compactness and access speed over
character flat files or SAS data sets. Also, it uses the logic
of the native CRSP character file in the way that it breaks
the CRSP data vectors into records and identifies them. This
allows a programmer to make use of character-CRSP techniques
with fairly minimal changes.

The eight FORTRAN files here, together with the CRSP-supplied
routines, support creation of and FORTRAN access to the fixed-
binary format.

These routines are simple modifications of the FORTRAN code
supplied by CRSP. They should work on Unix and OpenVMS systems
with only changes to the OPEN (and possibly INCLUDE) statements
in the main program.

bfstk.f is the stock file conversion program. It reads in an 
ascii character stock file using the original CRSP CHGET
subroutine and writes out the fixed-length binary file. The
comments in the program indicate where the user needs to make
changes. The program needs to be compiled and linked with the 
subprograms that it calls. On a Digital Unix system, this can
be done with the command:

% f77 -extend_source -o bfstk.o -O4 bfstk.f chget.f bical.f
 chckid.f ckdat.f cknum.f numrec.f outhdr.f bfput.f bidump.f

This puts the executable code in bfstk.o.

You will need to convert the stock index file also. The fixed-
binary stock index file is the same as the regular binary one,
so no new subroutine is needed. Just use the CRSP subroutine
CHCAL to read in the character index file and BICAL to write
out the binary file. However, make sure that the OPEN state-
ment for the binary index file sets it up as a fixed-length
(or stream_lf) unformatted file with an 11-word record length.
CRSP tends to define the binary index file as variable-length
even though all the records are in fact identical in length.

FORTRAN users can access the fixed binary CRSP file using
their familiar programs, but replacing some subprogram names
with those of the modified subprograms on this site. The
programs will have to be recompiled after the changes, of
course. The substitutions to make are:

Type		Original CRSP Character		Fixed Binary Replacement
include file	allincls			fallincl
include file	nmsfl				fnmsfl
subroutine	chget				bfget
subroutine	chgetn				bfgetn

The following are less commonly seen by the end user, but the indicated
substitution must be made when they are present in the user program.

Type		Original CRSP Character		Fixed Binary Replacement
include file	parmsfl				fparmfl
subroutine	chput				bfput
subroutine	chputn				bfputn

Comments, suggestions, bugs and questions may be sent to the author 
(mailto:arnie@iastate.edu). As there is more CRSP than Cowan in
the code files, any CRSP subscriber may use them freely except
for purpose of commercial sale of the code itself.

This document copyright (c) 1996 Arnold R. Cowan; all rights reserved.