Ramsey_model with inequality constraints gives wrong results?

Perfect foresight simulations
1

First I have 2 questions;

  1. should Dynare be able to solve the ramsey_model when there is an inequality constraint in the model (ie. the behaviour of the agents)?

  2. If it can, should this work with the perfect_forsight_solver using the lmmcp?

I tried this setup, but the path is quite different from the Social Planner setup, which I am confident is correct. So I want to know if Dynare should have obtained the correct solution with the ramsey_solver as well (and that a error in our code is the likely cause).

2

My hunch is that it should work in a perfect foresight context, but it may depend on the exact model specification. Could you maybe provide me with more details or the code to check what is going on?

3

Here’s the code that can replicate my results. Just run the ‘run_SP_vs_Ramsey’, and then you can plot tau with ‘plot_variants’. When there is no inequality constraint the solutions are the same. With the inequality constraint (NNCON definition) the solutions differ widely. And I dont see how the Ramsey solution should be different from the Social Planner solution in this case.

Thank you for you help :slight_smile:

SP_vs_Ramsey_inequality.zip (3.7 KB)

4

My hunch is that the mcp tag is not properly taken into account for the Ramsey FOC. Have you tried a brute force approach of using a max() operator for the investment constraint instead to see whether the problem persists?