Forecast with calibrated model - non-linear and linearized versions

Perfect foresight simulations
1

Hi there,

Can someone tell me how to do the following forecasts using a calibrated model?

  1. Forecast using non-linear model, and combining perfect foresight and surprise shocks with conditioning on exogenous and endogenous variables? I suspect it may be done by the function ‘det_cond_forecast’ but the documentation to this function does not explicitly say that the simulation is non-linear (the other related function - conditional_forecast - explicitly says it works with first order approximation).

  2. Using linear approximation of the model, and combining perfect foresight and surprise shocks with conditioning on exogenous and endogenous variables? For this it seems the ‘conditional_forecast’ in combination with conditional_forecast_paths may do the trick, but from the documentation it seems it supports only surprise shocks and not perfect foresight (or combination of both).

Many thanks for your answers.

Adam

2

det_cond_forecast belongs to the extended_path method and therefore is nonlinear, but deterministic. The conditional_forecast command in contrast is based on a linear state space model. It does indeed not handle anticipated shocks (for now). If your model and the anticipation horizon are not too big, you may be able to work around this restriction by introducing news shocks to handle the anticipation. At first order, certainty equivalence will assure that the distinction between a news shock and a deterministic anticipated shock is not important.

3

Johannes, thank you for the reply.

What do you mean by the “news shocks” exactly?

4

I am referring to the type of anticipated shocks as in Schmitt-Grohé/Uribe (2012): “What’s news in business cycles” or Jaimovich/Rebelo (2009): “Can news about the future drive the business cycle”