Knut Are Aastveit, Andrea Carriero, Todd E. Clark, and Massimiliano Marcellino, "Have Standard VARs Remained Stable Since the Crisis?" Journal of Applied Econometrics, Vol. 32, No. 5, 2017, pp. 931-951. This file details the data and computer programs used to produce the results in the published paper. Contact information: todd.clark [AT] clev.frb.org, (216) 579-2015. Most files are ASCII text files in DOS format. These are zipped in the file accm-files.zip, which has several folders. Unix/Linux users should use "unzip -a". The Excel (.xls) and PDF (.pdf) files are zipped in the file accm-binary.zip. Unix/Linux users should *not* use the "-a" option when unzipping this file. ******************** DIRECTORY STRUCTURE: ******************** "data" folder: files of time series data used in RATS programs detailed below "procedures" folder: computer programs in RATS for model estimation and forecasting, called by other programs "programs" folder: computer programs in RATS that read in and transform data, call procedures, and produce the results provided in the paper "RegimeSwitchingVARs" folder: files of data and Matlab programs for regime-switching results of paper ****************** DATA FOLDER FILES: ****************** The following files contain the time series of data used in the article (all at the quarterly frequency), with rows corresponding to dates and columns to variables. The first row lists the variables. The first column in each file provides the dates of the observation. Quarters 1, 2, 3, and 4 are labeled as months 1, 4, 7, and 10, respectively. We provide the data in both comma-separated value text files (those with extension .csv, in the folder "CSVfiles") and Excel files (those with extension .xls). mlydata.xls: monthly data, 7 variables qlydata.xls: quarterly data, 3 variables rtd.bfi.xls: quarterly real time data on business fixed investment, quarterly vintages starting in 1965:Q4 rtd.cum.xls: monthly real time data on capacity utilization in manufacturing, quarterly (months 2,5,8,&11) vintages starting in Feb.1985 rtd.gdp.xls: quarterly real time data on real GDP, quarterly vintages starting in 1965:Q4 rtd.ip.xls: monthly real time data on total industrial production, quarterly (months 2,5,8,&11) vintages starting in Feb. 1985 rtd.payrolls.xls: monthly real time data on payroll employment, quarterly (months 2,5,8,&11) vintages starting in Feb. 1985 rtd.pce.xls: quarterly real time data on PCE deflator (price index), quarterly vintages starting in 1965:Q4 rtd.pcex.xls: quarterly real time data on core PCE price index, quarterly vintages starting in 1996:Q1 rtd.rcon.xls: quarterly real time data on real consumption, quarterly vintages starting in 1965:Q4 rtd.resinv.xls: quarterly real time data on residential investment, quarterly vintages starting in 1965:Q4 rtd.unemprate.xls: monthly real time data on unemployment rate, quarterly vintages starting in 1965:Q4 SPF.gdp.xls: quarterly data on SPF nowcasts (current quarter forecasts) of GDP growth SPF.pcex.xls: quarterly data on SPF nowcasts (current quarter forecasts) of core PCE inflation SPF.tbill.xls: quarterly data on SPF nowcasts (current quarter forecasts) of 3-month T-bill rate SPF.unemprate.xls: quarterly data on SPF nowcasts (current quarter forecasts) of the unemployment rate Data sources: In our formal assessment of break probabilities, regime switches, and conditional forecasts, we use quarterly data for 1959:Q1-2015:Q2 obtained in early October 2015 from the Federal Reserve Board's FAME database. In the analysis of real-time out-of-sample forecasts, we use real time data vintages through 2015:Q3, obtained from the Federal Reserve Bank of Philadelphia's Real-Time Dataset for Macroeconomists (RTDSM). In the out-of-sample forecast analysis, real-time data are used for GDP, core PCE prices, consumption, business fixed investment, residential investment, industrial production, capacity utilization in manufacturing, non-farm payroll employment, and headline PCE inflation. For the other variables (unemployment, the federal funds rate, the credit spread, and stock prices), for which data are either not revised or only slightly revised, we rely on just currently available time series. In constructing forecasts at each point in time, we use only the data (for the model variables in use) that would have been available at the time the models would have been estimated and forecasts would have been constructed. ************************ PROCEDURES FOLDER FILES: ************************ Note: all files use the software programming language RATS. All VAR estimation uses Bayesian methods. BVAR.src: procedure for estimating and forecasting with conjugate Normal-Wishart prior and posterior BVAR.terrors.src: procedure for estimating and forecasting with VAR with fat tails BVARwithbreak.src: procedure for estimating with Bayesian methods VAR with discrete break BVARwithbreak.terrors.src: procedure for estimating with Bayesian methods VAR with discrete break, allowing fat tails in errors TVPSV.src: procedure for estimating and forecasting with VAR with time-varying parameters and stochastic volatility TVPKK.src: procedure for estimating and forecasting with VAR with time-varying parameters and volatility, under the Koop-Korobilis approach conditforecastbyKF.src: procedure for producing conditional forecasts using a Kalman Filter-based approach, given draws of parameters fcmoments.oos.src: procedure for processing draws of out-of-sample forecasts to obtain posterior moments of interest fcmoments.src: procedure for processing draws to obtain posterior moments of interest, used in conditional forecast exercise ********************** PROGRAMS FOLDER FILES: ********************** NOTES: (1) For out of sample forecasting, to facilitate use by other researchers, the code we have provided incorporates some simplifications relative to the code actually used to produce the results in the final version of the paper. The primary simplification is the shortening of the sample to the one used in the final paper. This shortening of the sample makes it practically impossible to match up the random number generator's seed values to the ones underlying the results in the paper. Accordingly, running the out-of-sample forecast programs we have provided will produce results close to those in the paper, but not exactly the same due to sampling variation in the MCMC estimation. For those interested in exactly replicating the paper's out-of-sample forecast results, we can provide the much larger set of files necessary to do so (the set is larger because, for a longer sample also considered in an earlier version of the paper, we broke the forecast production into pieces, with separate programs for each two year period, and then used additional programs to combine and compile the results). (2) The programs listed below read in and transform as needed the data and call the procedure files to produce estimates and forecasts as follows. (3) Some of these programs take a long time to complete (if producing 5000 forecasting draws; for development or experimentation, reducing the number of draws speeds calculation considerably). Table1_BVAR.prg: Produces Table 1's discrete break results for the baseline BVAR Table1_BVARfattails_baseprior_5df.prg: Produces Table 1's discrete break results for the BVAR with fat tails, the base prior, and 5 degrees of freedom in the t-distribution Table1_BVARfattails_baseprior_10df.prg: Produces Table 1's discrete break results for the BVAR with fat tails, the base prior, and 10 degrees of freedom in the t-distribution Table1_BVARfattails_altprior_10df.prg: Produces Table 1's discrete break results for the BVAR with fat tails, the alternative prior, and 5 degrees of freedom in the t-distribution Table1_BVARfattails_altprior_5df.prg: Produces Table 1's discrete break results for the BVAR with fat tails, the alternative prior, and 10 degrees of freedom in the t-distribution Figure1.prg: Produces Figure 1's conditional forecast results Figure2.prg: Produces Figure 2's conditional forecast results Figure3.prg: Produces Figure 3's conditional forecast results Figure4.prg: Produces Figure 4's conditional forecast results oosfc_VAR4.prg: Produces real-time OOS forecasts, BVAR, in 4 variables: (1) recursive, 1985 start and (2) rolling, 80 obs. window oosfc_VAR4_avgwindow.prg: Produces real-time OOS forecasts, BVAR, in 4 variables: (3) avg. rolling window oosfc_VAR4_TVPSV.prg: Produces real-time OOS forecasts, BVAR, in 4 variables: (4) TVP-SV oosfc_VAR13.prg: Produces real-time OOS forecasts, BVAR, in 13 variables: (1) recursive, 1985 start and (2) rolling, 80 obs. window oosfc_VAR13_avgwindow.prg: Produces real-time OOS forecasts, BVAR, in 13 variables: (3) avg. rolling window oosfc_VAR13_TVPKK.prg: Produces real-time OOS forecasts, BVAR, in 13 variables: (4) TVP-KK oosfc_VAR4hybrid.prg: Produces real-time OOS forecasts, BVAR using SPF nowcasts as conditions, in 4 variables: (1) recursive, 1985 start and (2) rolling, 80 obs. window oosfc_VAR4hybrid_avgwindow.prg: Produces real-time OOS forecasts, BVAR using SPF nowcasts as conditions, in 4 variables: (3) avg. rolling window oosfc_VAR4hybrid_TVPSV.prg: Produces real-time OOS forecasts, BVAR using SPF nowcasts as conditions, in 4 variables: (4) TVP-SV ********************************* RegimeSwitchingVARs FOLDER FILES: ********************************* Data: Data files are provided as ASCII files in DOS format (extensions .csv) and as Excel files (extensions .xlsx). The file Data_USA contains data for 4 US macroeconomic variables for the data sample 1960:Q1-2015:Q2. GDP: Quarterly GDP growth (annualized) UNRATE: Civilian Unemployment Rate Inflation: Core inflation. Personal consumption expenditures (PCE) prices excluding food and energy prices Interest rate: The effective Federal runds rate Replication codes: The following MATLAB files replicates the results for the regime-switching VARs in the paper: run2regimes_Coeff.m: Code for the regime-swtiching VAR with breaks in VAR coefficients, 2 regimes (referred to as RS-C). The code replicates the results for RS-C in Table 2 in the main text and Figure 3 in the supplementary appendix run2regimes_Var.m: Code for the regime-swtiching VAR with breaks in the error variance, 2 regimes (referred to as RS-V). The code replicates the results for RS-V in Table 2 in the main text and Figure 4 in the supplementary appendix run2regimes_Independent.m: Code for the regime-swtiching VAR which allows for independent breaks in the VAR coefficients and the error variance, 2 regimes (referred to as RS-CV). The code replicates the results for RS-CV in Table 2 in the main text and Figures 5 and 6 in the supplementary appendix