MATLAB code to estimate and forecast the model discussed in the Liberty Street Economics blog post "The FRBNY DSGE Model Forecast".
All you need to do is run the file Main.m
. This script will run the entire
set of code, calling
set_paths.m
: Set default directories for input and output; add code subfolders to MATLAB path.spec_990.m
: Set model specifications and important flags for estimation and forecasting.gibb_est_ant.m
: Find posterior mode and sample from posterior distribution.forecast_parallel_est_ant.m
: Forecast observables; can run in parallel.forplot.m
: Load forecasts into data structures to prepare for plotting.plotPresentation.m
: Plot forecasts.
If you would like to change defaults for estimation and forecasts, see
spec_990.m
. There, you can modify
- Estimation Parameters
reoptimize
: Whether to re-optimize and find the mode or use saved mode.CH
: Whether to re-compute the hession or use saved.nsim
: The number of posterior draws per block.nblocks
: The number of blocks.nburn
: Size of the burn-in.jstep
: From the blocks, the forecasting code will only use every jstep-th element.
- Forecast Parameters
zerobound
: Whether to incorporate anticipated policy shocks.peachflag
: Whether to condition time T 1 forecasts of observables on user-provided forecasts of observables (treating the information supplied by the user as data).distr
: Flag to specify whether to parallelize the forecast procedure.nMaxWorkers
: Number of workers to use in parallel forecast procedure.
Some common issues that may arise while running with default settings include:
-
Can't load modal parameters: Check that you're running
Main
in the base (DSGE-2015-Apr
) directory. Our code uses relative paths throughout, including to specify the location of the mode file (save/mode_in
), so it won't be found if you're in a subdirectory. -
Negative diagonal element in Hessian: Make sure that you're reading in the provided mode file correctly. If you set
reoptimize = 1
and re-rancsminwel
before computing the Hessian, it's possible that you haven't found a true mode (perhaps because not enough iterations were used). -
Finally, see the section "Final Notes on MATLAB Versions and Toolboxes" below.
In the main folder, there exist the following directories to house code components:
data/
: Input datadsgesolv/
: Solving the model; includesgensys.m
code.estimation/
: Mode-finding and posterior sampling.figures/
: For output including parameter moment-tables and TeX tables.forecast/
: Forecasting programs.graphs/
: For graphs of forecasts.initializaization/
: Loading data, defining model structure, setting up important model flags.kalman/
: Kalman filtering and smoothing.plotting
: Loading forecast distributions and generating/saving plots.save/
: Input mode and output data generated by the code (output mode, posterior draws, forecasts).toolbox/
: Supporting programs.
This section describes important programs in greater detail. If the user is interested only in running the default model and reproducing the forecast results, this section can be ignored.
This section focuses on what the code does and why, while the code itself (including comments) provides detailed information regarding how these basic procedures are implemented.
Main Program: estimation/gibb_est_ant.m
Purpose: Finds modal parameter estimates and samples from posterior distribution.
Main Steps
- Initialization: Read in transform raw data from
data/
. Load files frominitialization/
related to model specification, parameter priors, parameter restrictions. - Find Mode: The main program will call the
csminwel.m
optimization routine to find modal parameter estimates. Can optionally start estimation from a starting parameter vector by specifyingdata/mode_in.
- Sample from Posterior: Posterior sampling begins from the computed mode,
first computing the Hessian matrix to scale the proposal distribution in the
Metropolis Hastings algorithm. Settings for the number of sampling blocks and
the size of those blocks can be specified in
spec_990.m
.
Remark: In addition to saving each draw of the parameter vector, the
estimation program also saves the resulting posterior value and transition
equation matrices implied by each draw of the parameter vector. This is to save
time in the forecasting step since that code can avoid recomputing those
matrices. In addition, to save space, all files in save/
are binary files.
Main Program: forecast/forecast_parallel_est_ant.m
Purpose: Compute forecast distribution for the observables, sampling from the full posterior distribution of parameters and sampling exogenous shocks.
Main Steps
- Load Draws: Load in posterior distribution blocks that are output from the estimation stage.
- Filter and Smooth: Pass matrices defining the state transition equation
into
forecastFcn_est_ant.m
, which will filter and smooth the states over the history. - Forecast: Compute forecasts using
getForecast.m
, which takes matrices corresponding to a posterior draw and uses them to iterate on the time T state vector to obtain forecasts, adding in draws of exogenous shocks as well.
Remark: The code can be run in parallel or sequentially. Running in parallel requires the MATLAB parallel toolbox.
Main Programs: plotting/forplot.m
, plotting/plotPresentation.m
Purpose: Generate plots of observables.
Main Steps
- Load Plots: The program
forplot.m
will load output from the forecast program into the workspace, computing entries in the data structuresMeans
andBands
. - Plot: The program
plotPresentation.m
will plot the forecasts and shock decompositions, saving the graphs in thegraphs/
folder.
Remark: Each time forplot.m
is called, it will attempt to recompute the
forecast means and bands, which takes some time. However, these are saved in
save/
after the very first call to forplot.m
. To use these saved Means and
Bands structures and avoid recomputing, set useSavedMB=1
before running
forplot.m
.
In certain functions implemented in this program are Toolbox functions provided
by Mathworks. If you are receiving errors in running these programs due to
undefined functions, it is likely because you do not yet have access to these
Toolboxes. For example, to run the forecasts in parallel (the default setting),
you will need the parallel toolbox. Also, dlyap.m
, a function to solve
discrete-time Lyapunov equations, comes from the Control System Toolbox.
These programs are meant to be run in Matlab09a. While we have not attempted to run these programs using a more recent version of Matlab, it is possible that some of the Matlab-defined functions are not identical and thus may yield nonidentical results.
Copyright Federal Reserve Bank of New York. You may reproduce, use, modify, make derivative works of, and distribute and this code in whole or in part so long as you keep this notice in the documentation associated with any distributed works. Neither the name of the Federal Reserve Bank of New York (FRBNY) nor the names of any of the authors may be used to endorse or promote works derived from this code without prior written permission. Portions of the code attributed to third parties are subject to applicable third party licenses and rights. By your use of this code you accept this license and any applicable third party license.
THIS CODE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT ANY WARRANTIES OR CONDITIONS OF ANY KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY WARRANTIES OR CONDITIONS OF TITLE, NON-INFRINGEMENT, MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, EXCEPT TO THE EXTENT THAT THESE DISCLAIMERS ARE HELD TO BE LEGALLY INVALID. FRBNY IS NOT, UNDER ANY CIRCUMSTANCES, LIABLE TO YOU FOR DAMAGES OF ANY KIND ARISING OUT OF OR IN CONNECTION WITH USE OF OR INABILITY TO USE THE CODE, INCLUDING, BUT NOT LIMITED TO DIRECT, INDIRECT, INCIDENTAL, CONSEQUENTIAL, PUNITIVE, SPECIAL OR EXEMPLARY DAMAGES, WHETHER BASED ON BREACH OF CONTRACT, BREACH OF WARRANTY, TORT OR OTHER LEGAL OR EQUITABLE THEORY, EVEN IF FRBNY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES OR LOSS AND REGARDLESS OF WHETHER SUCH DAMAGES OR LOSS IS FORESEEABLE.