Dear Professor Kao,
    I have all your programs running - OLS, FMOLS and DOLS as well as the
residual based LM test for FM and DOLS. Using several different data sets,
when I ran the FM, results showed stationary or near stationary residuals.
However, when I ran the DOLS, the LM stat was very or very, very large,
indicating rejection of cointegration. I think the reason is that the ps
program that runs the DOLS program in COINT 2.0 (in cregrs.src) reports the
incorrect sigma (standard error of the residuals).  The program does not
take the square root; whereas, other programs in cregrs.scr such as the FM
do take the square root (like it should) in calculating the standard error
of the residuals. When I took the square root, my results were much more
favorable and similar to the FM results.   Hence, the reason for the large
rejection of stationary residuals for DOLS in your paper "How does your
Output Grow?..." may be due to the COINT programming error.  
Sincerely,
Jack Strauss
Sang Sup Cho