Version 3

Changes in Dynare version 3

Version 3.06


Bug fixed:

  • bug with conditional compilation instructions in MODEL block
  • bug with variance computation


Bug fixed:

  • bug with penalty in Metropolis iterations
  • new version of Chris Sims optimizer (thanks Chris!)
  • bug in osr
  • bug in estimating correlation coefficients
  • typos in metropolis.m
  • function call with empty arguments that caused problems with Matlab version < 7


New features:

  • optimal simple policy rules and optimal policy under commitment for linear models
  • forecasting for calibrated models
  • anticipated exogenous deterministic shocks in the simulation of stochastic models
  • conditional compilation of part of the *.mod file

The manual has been updated (the pdf version is deficient, use the HTML one)

Bug fixed:

  • bug in Metropolis iterations with one chain
  • removed restriction on name of observed variables
  • Dynare uses now 500 draws among Metropolis iterations when computing posterior distribution of forecast, IRF, moments or smoothed variables
  • Corrected a bug in Diffuse Kalman filter and smoother

Version 3.05


  • correcting bug in computation of 2nd order approximation for models with leads on more than one period. Please redo your computations. Sorry about that.
  • correcting bug in estimation of models with lags on more than one period
  • correcting bug in forecasting after estimation
  • correcting bug in relative_irf in estimation
  • binaries for Linux x86 are released

Version 3.04


  • correcting bug in fs2000a example
  • correcting several bugs in posterior distribution of various statistics
  • correcting inefficiency in simulation of second order approximated models
  • correcting the bug with a parameter called 'c' in deterministic models


  • correcting bug with option optim in estimation


  • correcting bug with option bayesian_irf in estimation


  • correcting bug for 2nd order approximation with leads on more than one period
  • correcting bug for backward-looking models with lags on more than one period
  • correcting bug for forecast of trending variables in estimation
  • correcting bug in computation of modified harmonic mean estimator of marginal density in estimation
  • correcting bugs and modified diffuse filter and smoother (thanks to Marco Ratto for suggestions)
  • adding a noprint option in stoch_simul for better loops
  • an expression can be used to set option simul_seed
  • from now on, models which don't have a unique stable solution (Blanchard and Kahn condition) are never computed by stoch_simul

The manual still needs to be updated

Version 3.03

  • correcting bug on computation of (correction) in 2nd order approximation present in version 3.02 and 3.021
  • correcting bug on check with backward models
  • newer version of the optimizer contributed by Marco Ratto (mode_compute=5)

Version 3.02


  • correcting bug with deterministic shocks
  • correcting bug with check for backward models
  • correcting bug in creation of TeX tables in dynare_estimation (yet undocumented feature)


  • allows for leads on more than one period for 2nd order approximation
  • estimation of correlation coefficients of shocks or measurement errors are again possible (the manual still needs to be updated)
  • correction of bugs reported by M. Ratto, E. Pytlarczyk and JG Sahuc in the Diffuse Filter (thank you)

Version 3.01


Bug corrections to the Diffuse Smoother.


We use now the true diffuse prior Kalman filter proposed by Durbin and Kalman for nonstationary models.

Version 3.0

Version 3.0 is the first stable version with estimation support. It corresponds to the test version that was available through 2004. Note that the possibility of estimating correlation coefficients between shocks isn't available. There was a problem with the testing version to insure that the variance matrix of the shocks is indeed positive definite. Users should explicitly model that a given equation is affected by two uncorrelated shocks.