%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% readme file for:
%
% Kliem and Kriwoluzky: "Toward a Taylor Rule for Fiscal Policy"
%
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1) general

   All results are calculated with the DYNARE package. For the calculation of the main results 
   of the paper we modified the code slightly. Given the continously ongoing development of DYNARE, we 
   cannot ensure that all our adjustments run with all DYNARE distributions. For this reason, 
   we add the DYNARE package 4.2.4 to the replication files to ensure that all our results 
   are replicatable. Hence, for replication of the results add the directory:

   "dynare4.2.4/matlab" 

   to your matlab path.

2) subdirectory: data

   This directory contains a *.m file with the data used for estimiating the models in the paper. See the Appendix for a
   detailed discription of the data.


3) subdirectory: estimation

   In this directory you find the *.mod files to replicate the estimation of the baseline model (model_baseline.mod)
   as well as recommended model (model_ext.mod). The reported results in the paper are based on the posterior distribution
   approximated using a Metropolis Hastings Algorithm with two chains, each with 1Mio draws. Because 
   of storage space constraints, we do not offer all the draws. However, we put the results for the posterior mode as well 
   as the summary statistics in the directory (filenames extended by _backup). The figures w.r.t. posterior and prior distribution
   as well as a corresponding table can be generated by running:  generate_tables_figures_est.m

4) subdirectory: ramsey policy

   In this directory you find the files to do the steps explained in section 4.2 in the paper. 

	4.1 Run `dynare model_optimal' to simulate time series from the Ramsey solution of the model.
            At some point you will be asked if you want to load some specific file for the shock processes. Just press any key 
            to continue. Finally, the program saves the files: simul_final.mat and exo_simul.mat

        4.2. Now, run `dynare model_nonlinear' to estimate the optimized tax rules. We have putted our posterior mode estimates 
             into the directory (model_nonlinear_mode_backup.mat)
             Again, at some point you will be asked if you want to load some specific file for the shock processes. Now enter: `exo_simul.mat?
	     The program shows you a figure (as in the appendix) which compares the simulated time series from the Ramsey model and
             the model with optimized rules. The approximation of the posterior distribution at the posterior mode will be used next.           


5) subdirectory: identification

   In this directory you find the model_bench.mod files to replicate the identification analysis of section 4.3. in the paper. The program uses
   a modified version of the identification toolbox from DYNARE. The calculation in the paper uses 10k draws from the posterior distribution as
   approximated before (see point 4 in this readme or section 4.2 in the paper). (The definition of the marginal posterior for each parameter is
   hard coded!). Finally, the code saves the `sensitivity_results.mat?. This file can be used with `plt_identplots2.m? to generate the figure 2 
   in the paper as well as Table 5 in the Appendix. As before, we put our origianl result `sensitivity_results_backup.mat? into the directory. 

6) subdirectory: Second_order

   In this directory you find all codes necessary to minimzing the welfare cost in consumption-equivalent units. Finally, running 
   `dynare model_optimal? delivers Table 1 in the paper (section 5.1)

7) subdirectory: multiplier

   In this directory you find *.mod files for both models, baseline and recommended model (*_bench_irf1.mod and *_ext_irf1.mod). 
   Additionally, we put matrices with 500 posterior parameter vector draws for each model into the directory. By running:

    generate_table_multiplier.m

   the program automatically starts the baseline model, loads the posterior draws, and calculates the multplier as described in 
   section 5.3 in the paper. Afterwards, the steps are done for the recommended model too. Finally, the program prints a Table: Multiplier
   as in the paper.