Macro Modeling at the FRB: FRB/US

Today we take a look the Fed's key tool for economic modeling and forecasting, the large-scale semi-structural FRB/US model.

The FRB/US Model: A Tool for Macroeconomic Policy Analysis

Federal Reserve Board has a long history of serious commitment to the development, improvement and use of large-scale structural macroeconometric models. As most of you know, the current version of the model in use at the Board is the so-called FRB/US model–a forward-looking structural model of the U.S. economy with about 60 stochastic equations and 300 identities.

Macro Modeling at the FRB: FRB/US

FRB/US is still the main macroeconomic model in use at the Board.

(Some) current uses

• Forecasting, short and long-term

• Alternative scenarios

• Forecast confidence bands (stochastic simulations)

• Optimal monetary policy simulations

• Stress tests

The Fed makes this tool available to everyone here. It is written in Python, and some of the dependencies only run on Linux. It is a bit difficult to set up the FRB/US simulator to run in Windows, but our team did just that and tested this sophisticated modeling software in Windows 11 with WSL.

The modeling assumptions are very complex, well documented, and are located in a markup file "..\models\model.xml" for the simulation to use. To begin to modify it in meaningful ways, it takes at least a master's degree in economics - so we will trust the modeling assumptions.

FRB/US Model, Windows 11 with WSL (Ubuntu).

FRB/US Model: Brayton, Laubach, and Reifschneider - Stochastic Simulation.

Input: The historical FRB/US database (csv format) and FRB/US baseline databases (csv format and EViews edb format). The baseline merges historical values with a projection whose initial part follows the median path in FOMC's Summary of Economic Projections (SEP). Beyond the horizon of the SEP, the baseline projection is a model extrapolation.

Output: Predictions of Core PCI, Fed Funds rate, GDP growth, and unemployment rate.

Model uses bout 40 stochastic equations and 300 identities. It takes almost 1 minute to run on a decent laptop:

• Processor Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz, 2592 Mhz, 6 Core(s), 12 Logical Processor(s)

• Installed Physical Memory (RAM) 32.0 GB

Some key observations:

• Clearly, the unemployment model falls apart at about 2024Q1, since the band after that period is too wide to to be significant.

• Interestingly, nearly all models agree that inflation should return to pre-covid levels towards the end of 2023.

• Note: the data for these calculations goes through April 21, 2022, so this model is missing pertinent inflation data for May, which is rocking the markets this week, ahead of a key Fed interest rate hike in a few hours.

FRB/US Model - Stochastic Simulation

Exercising our due diligence, we Looked at some European models, and the OECD, and all of these mostly agree that inflation should be back down to pre-Covid, or near pre-Covid levels, in about 18 months.

Caveat: these models don't work well in "unprecedented" periods of economic upheaval, such as a complex, global recovery from a pandemic.

Example:

The New York Fed DSGE Model (Dynamic stochastic general equilibrium).

The DSGE model is much smaller than the FRB/US.

The charts below show the New York Fed's DSGE forecast for December 2021, overlayed on top of their previous estimate.

The NY Fed missed in spectacular fashion in their forecast of Core PCE Inflation for

Dec 2021:

Old forecast: 2.26

New Forecast: 5.00

This major discrepancy was the genesis of our team getting into the analysis of the Fed's tools for economic modeling and predictions.

Using the FRB/US modeling simulator, we can see how the Federal Open Market Committee (FOMC) meeting participants obtain their estimates and projections:

In conjunction with the Federal Open Market Committee (FOMC) meeting held on

December 14–15, 2021, meeting participants submitted their projections of the most

likely outcomes for real gross domestic product (GDP) growth, the unemployment

rate, and inflation for each year from 2021 to 2024 and over the longer run.

We also took in this excellent scholarly paper by Ben Bernanke for the Brookings Institution.

Bernanke discusses in detail the FRB/US modeling in 2020

The New Tools of Monetary Policy American Economic Association Presidential Address Ben S. Bernanke for Brookings, January 4, 2020

The next FOMC meeting (and Fed rate hike) is in 18 hours.

A 0.5% rate hike is expected, a 0.75% rate hike is being discussed in the media.

CME countdown to FOMC meeting.

Update: CNBC - Fed hikes its benchmark interest rate by 0.75 percentage point, the biggest increase since 1994