This paper analyzes the relation between forecasts and past and current information, with special attention to the effects of policy interventions. The paper defines impact factors as the multiplier of changes in information variables in the total effect of policy measures. Impact factors (IF) computed for vector autoregressive processes (VAR) integrated of order 1 and 2, (I(1) I(2)). IFs are shown to depend on the moving average impact matrices of the stationary representation of the process and on other parameters. The relation of impact factors with total multipliers and impulse responses is analyzed. Estimation of impact factors is discussed.