Numerical differences

Dear all!

I have a model, in which the blanchard-kahn conditions are fulfilled, the steady state of all log-linearized variables is zero and I get irfs.
My model defines an equation for capital accumulation based on past investment and the rental rate of capital is simply r_t=-k_t. My profits are defined as \pi_t=r_t+k_t. What I do not understand is that the time path of profits differs somewhat from zero, although I thought the sum of r_t and k_t should be exactly zero.
Does this have to do with restrictions from other equations or with first-order approximation in dynare?

timepathk_r.pdf (3.9 KB)
timepathprofit.pdf (3.5 KB)

With kind regards

The difference is of size 10^{-16}. That is well below machine precision and nothing to worry about. This happens when you work with binary representations of exact numbers.