Anticipated news shock for TFP

If we want to get IRFs to the positive news shock about future TFP, then in Dynare it is fairly easy to simulate the news shock. Supposed that original TFP shock is given by
At = rho_A*At(-1) + eps_A.
In order to simulate the news shock, I write two dynare codes as:

  1. A = news(-4); news = news(-4) + eps_A; where A and news are specified in var-command, eps_A is specified in varexo-command, and std(eps_A) = 0.01;
  2. A = rho_A*A(-1)+eps_A_news(-4); where A is specified in var-command, eps_A_news is specified in varexo-command, and its std is 0.01.
    What is correct?

The second. See