Interpretation of log transformed first differences in SVAR models

I am working with a structural VAR analysis, and want to understand my results. The model studies how a shock on one variable affects a second variable. I understand that if the first month's response is +3 %, it means that the initial value of the response variable increases by 3 percent following a shock. However, I'm uncertain on how to interpret the second month. If the value for the second month is +2%, is the accumulated effect 1,03*1,02 = approx. 5 % increase from the initial value? Or does this mean that the initial response falls from 3 % to 2 % in the second month, leaving the overall effect at 2%?