Declaration order of variables and equations

[rmauad]

Dear Professor Dennis,

I work at the Research Department of the Central Bank of Brazil, and our team has been using the SolveDSGE package for Julia. I am contacting you to confirm some findings about the package.

Using the attached example model (“test_kf.txt”), we found that the declaration of variables, shocks, and equations must follow a specific pattern to ensure the order of the coefficients in the solution matrices H and K are consistent. In the attached file “Declaration order of variables and equations.docx”, I describe this pattern which depends on the endogenous or exogenous nature of the variables. In the example model, we have two exogenous variables me_X and Y, and one endogenous: XX – which depends on Y.

By breaking each of the steps described in this file, the order of the coefficients in the solution matrix H becomes inconsistent with those in matrix K. Please see the attached “Solution matrices.docx” file.

Could you please confirm if our understanding is correct? If so, may I suggest that you include a “declaration order” section in the package manual?

Thank you in advance.
Best regards,
Roberto Baltieri Mauad
[Declaration order of variables and equations.docx]
(https://github.com/RJDennis/SolveDSGE.jl/files/13599985/Declaration.order.of.variables.and.equations.docx)
Solution matrices.docx
Archive.zip

[RJDennis]

Hi Roberto,

I have received and read your comment, but I haven't had time to look into it. I believe I know what you are referring to, however, but I'll wait until I have a chance to look into it before I respond.

Richard

[rmauad]

No problem, please take your time. Thank you!

[paulbousquet]

I just encountered a similar issue -- an error was thrown that the shocks weren't AR(1). When I printed the relevant line of code in solution_functions.jl, the matrix being checked to be a diagonal was [0 1;1 0]. My shocks were listed as e_r, e_a in the shocks block, but in the equations block the equation for A (containing e_a) was entered first. Once I switched them, this was no longer a problem.