We tested and ran our program on UNIX machines DECStations Sun Ultra Enterprise
5000, Enterprise 10000 and 60 Elite 3D with both operating system Solaris 2.5.1
and Solaris 2.6.

How to proceed:

1) Unzip the file kb.zip. At the Unix prompt run the command
	
	unzip kb.zip

   The files: kb.f 
              kbnotes.ps
              tensolve.f 
              blas.f
              uncmin.f 
              menus.m
              inputs.m
              TENSOLVE_DOC.ps
              PAR1,PAR2,...,PAR15
              start
              restart
              TESTRE001.RES
              TESTRE001.DAT
              TESTRE
              TESTRB001.RES
              TESTRB001.DAT
              TESTRB
   will be then extracted.

2) Compile the fortran files to get an executable file 'name.out'. On a 
   Unix machine at the Unix prompt run the command

	f77 -o name.out *.f

3) Test the program by running the command

	./name.out < start

   You should get the Rational Expectations results in the files
   REE001001.RES, REE001 and REE001001.DAT which should report the same 
   results as in the files TESTRE001.RES, TESTRE and TESTRE001.DAT.
   Modify the input parameters lambda and mu and set both of them equal
   to [1.1 1.1 1.1 1.1 1.1 1.1 1.1 1.1]. To modify these parameters follow
   step 5. Then, at the Unix prompt run the command

	./name.out < restart

   You should get the Rational Beliefs Equilibrium results in the files
   RBE001001.RES, RBE001 and RBE001001.DAT which should report the same 
   results as in the files TESTRB001.RES, TESTRB and TESTRB001.DAT.

4) Read the document kbnotes.ps and proceed with your own calculations.
   You can simply run the command

	./name.out

   and input the initialization variables in 'interactive mode'.
   Alternatively, you can set the initialization variables in the ascii 
   file 'start' when solving for a new case with uniform starting values, 
   or in the ascii file 'restart' when solving for a new case using as the
   initial guess the solution of a previous case, and then run one of 
   the commands used above:

	./name.out < start
	./name.out < restart

4.a)  Note that the files 'start' and 'restart' have been set with the 
      following initialization values for the TENSOLVE procedure
      (by input order):
   
	MSG                           =  16
	ITERATION LIMIT               = 150
	STEPMAX                       =   1.0d+03
   	FUNCTION TOLERANCE            =   1.0d-08
  	GRADIENT TOLERANCE            =   1.0d-27
	STEP TOLERANCE                =   0.0d+00
	RADIUS                        =  -1.0d+00
	METHOD (NEWTON)               =   0
	GLOBAL STRATEGY (LINE SEARCH) =   0.
   
   We refere you to the included postscript file TENSOLVE_DOC.ps for an 
   exhaustive explanation of the role played by the initialization 
   variables. You can set these variables to new values by modifying the 
   files 'start' and/or 'restart' by using any text editor. The list
   of values for the variables used in each case is reported at the
   top of the summary output file (see prologue of kb.f).

4.b) In the file 'start' to open output files and supply a uniform initial 
     guess to the non-linear equation solver the following input variables 
     follow those listed in 4.a (by input order):

	6 CHARACTER ROOT FOR OUTPUT FILENAMES   = 'REE001' 
	STARTING VALUE MENU OF CHOICES          =  1
	UNIFORM INITIAL GUESS OF STOCK PRICE    = 27.00d0
	UNIFORM INITIAL GUESS OF BOND PRICE     =  0.95d0
	UNIFORM INTTIAL GUESS OF STOCK HOLDINGS	=  0.50d0
	UNIFORM INITIAL GUESS OF BOND HOLDINGS  =  0.00d0

4.c) In the file 'restart' to open output files and supply the solution of a
     previous case as initial guess to the non-linear equation solver the 
     following input variables follow those listed in 4.a (by input order):

	6 CHARACTER ROOT FOR OUTPUT FILENAMES   = 'RBE001' 
	STARTING VALUE MENU OF CHOICES          =  2
	13 CHARACTER OLD DATA FILENAME          = 'REE001001.DAT'

4.d) Use any text editor to edit the files 'start' and/or 'restart' in order
     to modify the values of all the variables listed above so to fit the
     specifications of your own calculations.
  
5) To modify input parameters run matlab by typing 'matlab' at the Unix
   prompt. Then type 'inputs' when the matlab prompt appears. The following
   menu of choices will appear:

   ----- Input of parameters -----
 
        1) beta1
        2) beta2
        3) gamma1
        4) gamma2
        5) alpha1
        6) alpha2
        7) W1
        8) W2
        9) phi
        10) high-div
        11) low-div
        12) lambda
        13) mu
        14) a
        15) b
        16) exit
 
  Select a menu number:  
  
  and follow instructions.

  If matlab is not installed on your workstation, although more cumbersome,
  you can edit the PAR1,...,PAR15 files directly in any text editor, making 
  sure you have matched up the correct PAR file with its corresponding
  parameter as given in the menu above.


6) Of course many things could be improved in the code thus any suggestion
   will be greatly appreciated (please email mordecai@leland.stanford.edu).