Tolerance used by "solve_algo" to find ss from in

I find that all four “solve_algo” options give me a sligtly different ss solution depending on small changes on the initial conditions (with differences smaller than o.oo1). (I am supplying a initial values very clsoe to actual ss).

These small differences seem to matter for the correction of the constant term in the second order approximation. I.e., the corrections may be significantly different depending on the initial values used.

Is it possible to alter the tolerance used by the “solve_algo” to force the ss solution to be more accurate?

There is currently some confusion about the tolerance level used by the various solve_algo algorithms. I hope to be able to fix it in the coming months.
For the time being, you need to look at the code, starting with dynare_solve.m

I’m however surprised that it makes a big difference on the correction term.