DSGE.jl

The DSGE.jl package implements the New York Fed DSGE model and provides general code to estimate many user-specified DSGE models. The package is introduced in the Liberty Street Economics blog post The FRBNY DSGE Model Meets Julia.

This Julia-language implementation mirrors the MATLAB code included in the Liberty Street Economics blog post The FRBNY DSGE Model Forecast.

The New York Fed DSGE team is currently working on adding methods to solve nonlinear and heterogeneous agent DSGE models. Extensions of the DSGE model code may be released in the future at the discretion of the New York Fed.

Table of Contents

• Learning How to Use DSGE.jl
• Building New DSGE Models
• Estimation
• Forecasting
• Model Design
• Special Model Types
• The PoolModel Type
• DSGE-VARs and the DSGEVAR Type
• DSGE-VECMs and the DSGEVECM Type
• Auxiliary Methods for DSGE-VARs and DSGE-VECMs
• Model Implementation Details
• The AbstractDSGEModel Type
• Defining Indices
• Type Interfaces
• The PoolModel Type
• The DSGEVAR Type
• The DSGEVECM Type
• Running an Existing Model
• Running with Default Settings
• Input/Output Directory Structure
• Input Data
• Setup
• Loading data
• Non-FRED data sources
• Incorporate population forecasts
• Dataset creation implementation details
• Conditional data
• Common pitfalls
• Update sample input data
• Data Transforms and Utilities
• New York Fed DSGE Model 990 Data
• Data Series
• Interest Rate Expectations Data
• References
• Disclaimer
• Solving the Model
• The gensys routine
• Regime-Switching
• The System Type
• The RegimeSwitchingSystem Type
• The TimeVaryingInformationSetSystem Type
• Estimation
• Procedure
• Metropolis-Hastings Sampler
• SMC Sampler
• Forecasting
• Procedure
• Forecast Outputs
• Preparing Forecast Inputs
• Computing Forecast Outputs
• Saving Forecast Outputs
• Forecasting Functions
• Impulse Responses
• DSGE Impulse Responses
• VAR Impulse Responses
• DSGE-VAR Impulse Responses
• VECM Impulse Responses
• DSGE-VECM Impulse Responses
• Wrappers for Impulse Response Functions
• Docstrings
• Computing Means and Bands
• Procedure
• Weighted Averages of Full-Distribution Forecasts
• Functions for Calculating Means and Bands
• The MeansBands type
• Alternative Policies
• Procedure for Permanent Alternative Policies
• Procedure for Temporary Alternative Policies
• Alternative Policy Uncertainty and Imperfect Awareness
• MultiPeriodAltPolicy
• Types
• Alternative Scenarios
• Basic Scenarios
• Additional Features
• Aggregating Multiple Scenarios
• Forecast Decomposition
• Plotting
• Plotting Estimation Results
• Plotting Forecasts
• Other Plots
• Standard Algorithms
• Solving the Model
• Optimization
• Hessian Approximation
• Sampling
• State Space Filters and Smoothers
• Sequential Monte Carlo
• Advanced Usage
• Package Directory Structure
• Working with Settings
• Accelerating Computation of Regime-Switching System
• Regime-Switching Forecasts
• Alternative Policy Uncertainty and Imperfect Awareness
• Automatically Generating Anticipated Shocks
• Automatic Endogenous ZLB Enforcement as Temporary Rule
• Editing or Extending a Model
• Additional Tips
• Contributing to DSGE.jl
• Notes for DSGE.jl Contributors
• DSGE Julia Style Guide
• The DSGE MATLAB to Julia Transition: Improvements and Challenges
• Code Improvements
• Challenges
• Conclusion
• Disclaimer
• Macroeconomic Forecasting with DSGEs Using Julia and Parallel Computing
• Performance Improvements
• Design Principles
• Parallel Computing
• StateSpaceRoutines.jl
• Disclaimer
• References
• License

Acknowledgments

Developers of this package at the New York Fed include

• Michael Cai
• William Chen
• Shlok Goyal
• Abhi Gupta
• Alissa Johnson
• Pearl Li
• Ethan Matlin
• Erica Moszkowski
• Reca Sarfati
• Micah Smith

Contributors to this package at QuantEcon include

• Zac Cranko
• Spencer Lyon
• Pablo Winant

The gensys and csminwel routines DSGE.gensys and DSGE.csminwel are based on routines originally copyright Chris Sims. The files are released here with permission of Chris Sims under the BSD-3 License.

The kalman_filter routine is loosely based on a version of the Kalman filter algorithm originally copyright Federal Reserve Bank of Atlanta and written by Iskander Karibzhanov. The files are released here with permission of the Federal Reserve Bank of Atlanta under the BSD-3 License.