Welcome to the Macroeconomic Forecasting Platform

The Macroeconomic Forecasting Platform is a software platform running on top of Matlab (Dynare) for estimating and comparing the forecasting performance of various DSGE and time-series models. It features a graphical user interface for efficiently setting up estimating and forecasting options for the model you choose.

This documentation describes how to install and use the platform, and explains the output generated by the platform in details.

It also provides instructions on how to incorporate your own DSGE models into the platform, and how to compare the forecasting outcomes across various models.

If you have any questions, please send us an email at forecastingplatform@gmail.com

Major advantages

Rich models to estimate

The platform can generate forecasting results based on a wide range of estimated DSGE models and time-series models. It currently features DSGE models of various kinds. These include standard small-scale New-Keynesian model (e.g., Del Negro and Schorfheide, 2004), canonical medium-scale DSGE model (e.g., Smets and Wouters, 2007), as well as DSGE model with frictions on the financial sector (e.g., Del Negro et al., 2015) or housing sector (Kolasa and Rubaszek, 2015). It also offers two Bayesian VAR models with the classical Minnesota priors (Doan et al., 1984; Litterman, 1986) and the optimized GLP priors (Giannone et al., 2015).

Rich data to choose

The platform already includes the a rich set of macroeconomic data for estimation, which not only saves you the time for manually collecting data, it can also guarantee that the difference in models forecasts will never come from the difference in data collection and processing procedures.

There are more than thirteen observables for you to choose from, including:

  • Real output growth
  • Inflation measured by the GDP deflator
  • Federal Funds Rate
  • Consumption growth
  • Non-durables and services consumption growth
  • Durable goods consumption growth
  • Investment growth
  • Non-residential investment growth
  • Residential investment growth
  • Real wage growth
  • Hours worked
  • Spread on loans to firms: difference between the BAA corporate bond yield and US 10-year Treasury yield
  • Spread on mortgage loans: difference between the effective interest rate on conventional single-family mortgages and the Federal Funds Rate

These data ranges from 1960:Q1 to 2018:Q4, and it includes both real time data and revised data for different time vintages. Moreover, one could even augment the data by including the nowcast from the Survey of Professional Forecasters of the observables when estimating a model.

Rich options to specify

There are numerous of options to specify in model estimation and forecasts generation. The DSGE model can be estimated via either mode estimation or Metropolis-Hasting estimation techniques. You also have the freedom to decide data spans in different ways: by selecting the expanding series option, one simply needs to specify the date of the first observation, while by selecting the rolling window option, one simply needs to specify the length of the data.

Rich results to analyze

The platform will produce model forecasts based on the specific model and options you choose, and it will compare this result with forecasts generated by other models and options for respective vintages. In addition, it can also plot and compare impulse response functions based on the Bayesian IRF, as well as historical variance decompositions across various models.

User-friendly graphical interface

The graphical interface makes the above choices fast, simple and clear. You are just a few clicks away from generating your first forecasts!