When I run the attached .mod file Model2.mod, with a steady_state_model block, Dynare computes the steady state with no problem. When I replace the steady_state_model block with an initval block in file Model1.mod, with the exact values produced from Model2, I get a “trends not compatible with balanced growth path” error. I tried increasing the balanced_growth_test_tol parameter but that didn’t help. I am wondering why Model1 is getting this error.

The background is that I am working with a tiny generalization of a well-tested model, involving a parameter called “tau.” The original model assumes tau=1. I am trying to solve the model with tau<1. In that case I am getting a failure to find the steady state using steady_state_model (even though I can solve the steady state model directly). So I thought I would try using initval instead, but ran into this “trends not compatible” issue even with tau=1. Why the tau<1 case isn’t working is a deeper mystery, but I don’t expect help with that unless it’s something obvious. The two attached files both have tau=1. Model1.mod (6.3 KB) Model2.mod (6.5 KB)

@sebastien Do you know what is going on? Is my reading of RemovingTrends - DynareWiki correct that the evaluation point is always what is contained in initval and therefore a steady_state_model-block will not provide the evaluation point? That would suggest that there is indeed a problem with the BGP in the model.

I confirm that the BGP test in the preprocessor uses input from initval (or load_params_and_steady_state if present), but not from the steady_state_model block.

I am coming back to this question, because I still haven’t been able to figure out why I am getting this BGP error. As mentioned, when I have a steady_state_model block, the model seems to solve, and I can generate IRFs that look reasonable, as in Model2.mod (attached). When I instead use an initval block (as in Model1.mod), I get the BGP error. I understand from Sebastien’s answer that the BGP test relies on the initval block (using the numerical values of the steady state found from the steady_state_model), which you indicate suggests there’s a problem with the BGP in the model. But I can’t find the problem. Every equation seems balanced. I’m not asking you to debug the code, I’m just wondering if there’s some other explanation for why I’m getting this error. Or, to put it another way, I’m concerned that the results I’m getting when I use the steady_state_model block are somehow spurious. Thanks.
p.s. I tried changing the balanced_growth_test_tol, but that didn’t help. Also, note that the equation tagged by the error message is “chi=1/c” and the growth factors of chi and c are inverses. Model1.mod (3.9 KB) Model2.mod (5.5 KB)