+ Kaenzig data files - some superfluous files, update readme

.gitignore

@@ -5,3 +5,5 @@
 *.png
 *.eps
 *.DS_Store
+!applications/kaenzig/dataQuantM.mat
+!applications/kaenzig/OilSurprisesMLog.mat

README.md

@@ -3,7 +3,10 @@
 Matlab code for inference on variance decompositions and the degree of invertibility/recoverability in a general Structural Vector Moving Average (SVMA) model identified by external instruments (IVs, also known as proxies)
 
 **Reference:**
-Plagborg-Møller, Mikkel and Christian K. Wolf (2020), "Instrumental Variable Identification of Dynamic Variance Decompositions", https://scholar.princeton.edu/mikkelpm/decomp_iv (paper + online appendix)
+Plagborg-Møller, Mikkel and Christian K. Wolf (2021), "Instrumental Variable Identification of Dynamic Variance Decompositions", https://scholar.princeton.edu/mikkelpm/decomp_iv (paper + online appendix)
+
+**Acknowledgements:**
+We are grateful to Diego Känzig for allowing us to reproduce some of the data files used in his 2021 AER paper.
 
 Tested in: Matlab R2020a on Windows 10 PC (64-bit)
 
@@ -13,8 +16,9 @@ Tested in: Matlab R2020a on Windows 10 PC (64-bit)
 - [SVMAIV_estim.m](functions/SVMAIV_estim.m): main function for SVMA-IV inference
 - [SVARIV_estim.m](functions/SVARIV_estim.m): SVAR-IV inference (assumes invertibility)
 
-**[application](application):** empirical example
-- [run_gk.m](application/run_gk.m): example based on Gertler & Karadi (2015)
+**[applications](applications):** empirical applications
+- [gk/run_gk.m](applications/gk/run_gk.m): application based on Gertler & Karadi (2015)
+- [kaenzig/run_kaenzig.m](applications/kaenzig/run_kaenzig.m): application based on Känzig (2021)
 
 **[illustration](illustration):** numerical illustration
 - [run_sw.m](illustration/run_sw.m): SVMA-IV and SVAR-IV analysis of Smets & Wouters (2007) model
@@ -55,4 +59,12 @@ Parameter names:
 - `FVR`: forecast variance ratio
 - `FVD`: forecast variance decomposition
 
-See the [empirical application](application/run_gk.m) for a concrete example. Additional optional arguments to [`SMVAIV_estim.m`](functions/SVMAIV_estim.m) are listed at the top of the function.
+See our [empirical applications](applications) for concrete examples of how to use the code. Additional optional arguments to [`SMVAIV_estim.m`](functions/SVMAIV_estim.m) are listed at the top of the function.
+
+## Replication instructions
+
+The figures and tables in our paper and supplement are produced as follows:
+- Figures 1 and B.7 and Table 1: [applications/gk/run_gk.m](applications/gk/run_gk.m)
+- Table 2: First run [simulations/run_sim.m](simulations/run_sim.m) nine separate times, with the variable `model.dgp` at the top set to each of the options `0` through `8`. Then run [simulations/print_results.m](simulations/print_results.m)
+- Figures B.1-B.6 and Table B.1: [illustration/run_sw.m](illustration/run_sw.m)
+- Figures B.8-B.11: [applications/kaenzig/run_kaenzig.m](applications/kaenzig/run_kaenzig.m)

applications/gk/run_gk.m

@@ -32,7 +32,7 @@
 
 % Estimation settings (see other optional settings in "functions/SVMAIV_estim.m")
 settings = {'ic', 'aic';        % Information criterion
-            'n_boot', 500;      % Number of bootstrap samples
+            'n_boot', 1e3;      % Number of bootstrap samples
             'signif', 0.1;      % Significance level
             'horiz', 1:24}';    % Horizons of FVR to report
 

applications/kaenzig/run_kaenzig.m

@@ -4,68 +4,57 @@
 
 % Empirical application: Kaenzig (AER, 2021)
 
-%% Load data
-
-% settings
-
-data.endo_vars = {'oil_price','oil_production','oil_inventories','world_ip','neer','ip','cpi','ffr','vxo','tot'};
-data.iv        = 'oil_surprise';
-
-% data.smpl_start = '1974M01'; 
-% data.smpl_start = '1983M04'; 
-data.smpl_start = '1983M03'; 
-data.smpl_end   = '2017M12'; 
 
-% data
-
-load data_kaenzig
-
-% endogenous variables
+%% Load data
 
-data.sampleDates = sampleDates;
+% Settings
+data.endo_vars  = {'oil_price','oil_production','oil_inventories','world_ip','neer','ip','cpi','ffr','vxo','tot'}; % Abbreviations of endogenous variables
+data.diff       = [0,1,1,1,1,1,1,0,0,0]; % =1: take first differences of endogenous variable
+data.iv         = 'oil_surprise';
+data.smpl_start = '1983M03'; % Sample start point
+data.smpl_end   = '2017M12'; % Sample end point
 
+% Endogenous variables
+dat_endo = load('dataQuantM'); % Endogenous variables
+data.sampleDates = dat_endo.sampleDates;
 data.smplStartInd = find(strcmp(data.sampleDates,data.smpl_start));
 data.smplEndInd   = find(strcmp(data.sampleDates,data.smpl_end));
+data.Y = dat_endo.data(data.smplStartInd:data.smplEndInd,:);
 
-data.Y = dataEndo(data.smplStartInd:data.smplEndInd,:);
-
-data.diff = [0,1,1,1,1,1,1,0,0,0];
-
+% First differences
 for i = 1:size(data.Y,2)
     if data.diff(i) == 1
         data.Y(:,i) = [NaN;12 * (data.Y(2:end,i) - data.Y(1:end-1,i))];
     end
 end
 
 % IV
-
-proxyRaw = [oilProxiesWTIM(:,14)];
-
+dat_iv=load('OilSurprisesMLog'); % IV
+proxyRaw = dat_iv.oilProxiesWTIM(:,14);
 data.Z = proxyRaw(data.smplStartInd:data.smplEndInd,:);
 
-% adjust sample
-
+% Adjust sample
 if max(data.diff) == 1
     data.Y = data.Y(2:end,:);
     data.Z = data.Z(2:end,:);
 end
 
-clear sampleDates dataEndo oilProxiesWTIM proxyRaw
+clear sampleDates dataEndo oilProxiesWTIM proxyRaw;
+
 
 %% SVMA-IV inference
 
 disp('*** SVMA-IV analysis ***');
 
 % Preliminaries
 addpath('../../functions');     % Add folder with SVMA-IV analysis functions
-addpath('subroutines');         % Add folder with some application-specific auxiliary functions
 rng(2018);                      % Seed random number generator (for bootstrap)
 
 % Estimation settings (see other optional settings in "functions/SVMAIV_estim.m")
-settings = {'p', 12;        % Information criterion
-            'n_boot', 100;      % Number of bootstrap samples
+settings = {'p', 12;            % Lag length
+            'n_boot', 1e3;      % Number of bootstrap samples
             'signif', 0.1;      % Significance level
-            'horiz', 1:50}';    % Horizons of FVR to report
+            'horiz', 1:48}';    % Horizons of FVR to report
 
 % Run inference routines
 [bounds, id_recov, inv_test, settings_struct] = SVMAIV_estim(data.Y, data.Z, settings{:});
@@ -105,7 +94,7 @@
 
 % figure
 
-plots.xticks = 0:10:50; % X axis ticks for FVR plot
+plots.xticks = 0:12:48; % X axis ticks for FVR plot
 plots.titles = {'FVR of Oil Price', 'FVR of Oil Production Growth', 'FVR of Oil Inventories Growth', 'FVR of World IP Growth', 'FVR of U.S. NEER Growth', ...
     'FVR of U.S. IP Growth', 'FVR of U.S. CPI Growth', 'FVR of FFR', 'FVR of VXO', 'FVR of U.S. TOT'};
 plots.xlabel = 'Horizon (Months)'; % X axis label for FVR plot
@@ -131,11 +120,8 @@
 
 end
 
-clear i
-
-% table
+clear i;
 
-svmaiv_table
 
 %% SVAR-IV analysis for comparison (assumes invertibility)
 
@@ -165,8 +151,4 @@
 
 end
 
-clear i
-
-% FVD table
-
-svariv_table
+clear i;