Simulation results change even with fixed seed

I am running a dynare program where in the simulations, even though I fix the seed, if I run it twice, the results change. Before each run, I clear the memory. I’m talking about things like the standard deviation of variables. I ran it on my apple macbook pro, on my unversity’s research computing cluster, and my coauthor ran it on her windows computer. In each instance, the results change.

I am running a dynare program using Matlab version R2016b Dynare 4.3.3 on a macbook pro. My university’s research computing cluster has Matlab R2016a and Dynare 4.3.3.

This is very strange to me. Has anyone else ever had this problem? Any ideas/suggestions/help would be greatly appreciated.

How do you fix the seed? Also, please try the Dynare 4.5. Older version may not be compatible with the newer version to adress Matlab’s random number generator.

set_dynare_seed(82784) ;

Have you tried Dynare 4.5?