Is there anyway that I can specify that certain computed variables in the steady state should be positive? At the moment dynare gives me a negative value for a variable that I expected to be positive. Or is the only explanation that there is a mistake in the model?

Either there is a mistake in the model, or the model has several steady states and you don’t get the one you want. When there are several steady states STEADY may give different answers for different initial values.

The best is to first study the problem analyticaly if you can.

Thanks Michel. I suspect there might be more than one steady state since I’m only taking this model from another paper and it seems to work fine there. I’ll give it a shot with different initial values and see how it goes.