General problems encountered using dynare

Hi Prof.,

I have a general question regarding dynare and general solution process.

Right now I am solving a model with: 1. productivity shocks with large sigma; 2. several occasionally binding constraints. 3. many choice variables. Currently I know dynare Occbin toolkit, but it is unable to converge under if I increase the shock’s sigma. Note that the shock volatility is estimated from the data and I cannot reduce it.

Then I turn to global solutions, i.e. value function iteration, but as I have many states, it is extremely slow. What I am curious is: 1. is there a possible way that I can use VFI with very coarse grids, then interpolate? I have this general idea but the problem is that
a) the model has occasionally binding constraints, so Value functions would have holes at some grids;
b) should I interpolate during iterations or after obtaining the solution? Is there a general rule?
Any references would be greatly appreciated!

  1. If there possibly exists another solution that is free from the above concern?

Thanks.

This sounds like a problem that is not computationally feasible as is. If the algorithms dedicated to this problem like DynareOBC do not work, then you may need to alter the model. There are ways to interpolate and work with discontinuities in the context of VFI (see e.g. https://doi.org/10.1016/j.red.2010.03.001), but having many states typically is an insurmountable problem.

Hi Prof.,

Thanks for your kind reply, I will check the link you provided.

Best,
R