Shocks in stochastic simulations

You need to be systematic. Whether you need to have logs here depends on the type of feedback you. Is it the percentage point tax rate that reacts to absolute deviations of debt? Or is it the percent change in tax rate relative to its steady state that reacts the the percentage deviation of debt from its steady state? That will determine where to put exp or log. Note that I would not recommend general exp() substitutions as there are better options. See Question about understanding irfs in dynare