Replication Files for "Evaluating Monetary Policy Counterfactuals: (When) Do We Need Structural Models?" by Caravello, McKay & Wolf

Tested in Matlab R2022b on a Dell Inspiron 15.

In order to produce most Figures and Tables, the posterior mode and posterior draws for policy causal effect matrices are required. Due to its size (~7.5 GB), these are not included in this repository. They must be downloaded from here. Then, the /suff_stats folder in this repository must be replaced with the one downloaded above.

To ensure that all codes run, the variable "path"---located near the top of the various m-files---needs to be changed to reflect the correct path in the user's machine.

readme.pdf: contains detailed instructions for using this repository. The main points are summarized below.

main_varplus.m: produces all figures and tables in the paper.

applications: creates Table 5.1 and Figures 6-9, D.1-D.2

• second_moments: creates Table 5.1 and Figures 6, 7, 8 and D.1-D.2
• hist_evol: creates Figure 7
• hist_scenario: creates Figures 8 and 9

invertibility: functions used to create Figure 1 of the paper.

model_estim: obtains the posterior for the causal effect matrices for all models, as well as the posterior model probabilities.

• get_param_post.m: obtain descriptive statistics for posterior of strucutral parameters.
• get_posterior_mode.m: obtains the posterior mode.
• get_posterior_probs.m: get posterior model probabilities, and create posterior draws across different subsets of models.
• plot_model_irfs.m: plot fitted impulse responses, as well as impulse responses for different models and horizons.
• sample_posterior.m: obtain posterior draws for each model.

var_inputs: estimate 1) Wold Decomposition, 2) IRF to monetary policy shock using VARs

• run_var_fcst_evol.m: creates VAR forecasts for each date, used when constructing the historical evolution.
• run_var_fcst_scenario.m: creates VAR forecasts at a given point in time, used when constructing the historical scenario.
• run_var_mbc.m: estimates the impulse response to the Main Business Cycle shock by Angeletos, Collard & Dellas (2020, AER)
• run_var_mp_adrr.m: estimates the impulse response to monetary policy shocks, using the shock series constructed by Aruoba & Drechsel (2024) and Romer & Romer (2004)
• run_var_spf_fcst_compare.m: compares the forecasting performance of VARs and the Survey of Profesional Forecasters.
• run_var_swfactors.m: compares the forecasting performance of VARs with and without the Stock & Watson (2016) factors.
• run_var_wold.m: estimates Wold IRFs for the sample 1960Q1 - 2019Q4
• run_var_wold_early.m: estimates Wold IRFs for the sample 1960Q1 - 2007Q1

_auxiliary_functions: auxiliary functions and routines.

• Our solution of the HANK model uses: first, several files from the replication codes of the article Ahn et al. (2017); second, the CompEcon toolbox of Miranda and Fackler, available here: www4.ncsu.edu/~pfackler/compecon; and third, the ergodicdist.m function, written by Marco Maffezzoli. The codes closely build on those used by one of the authors in Wolf (2024).
• The files jbfill.m and winsorize.m are available on Mathworks file exchange.
• The file regcyc.m is taken from the replication codes for the article Hamilton (2018).