Code

US GDP Nowcaster

usnowcast.net is an automated nowcaster for US GDP. The nowcaster uses a MIDAS approach for each main component of GDP and is updated daily. A hobby nowcaster that is not always fully maintained, with further details provided on the website itself.

DFM Nowcaster

A nowcasting dynamic factor model estimated using Bayesian methods, implemented in Python. As far as I know this is the only code available to implement a Bayesian version of a nowcasting DFM (most people use the EM algorithm made available from Giannone, Reichlin and Small 2008). The model can accommodate multiple factors, a choice of lags in the observation and transition equations, and can include an MA term in the errors in the observation equation. Also provides quasi out-of-sample mean squared errors to assess nowcasting performance. Roughly based on my work in Gauging the Globe: The Bank's Approach to Nowcasting World GDP .

SVAR Toolbox

A Matlab toolbox for running BVARs with long-run restrictions, the Max Share approach, and a standard Cholesky identification. Sign restrictions, zero restrictions, and forecast error variance magnitude restrictions have also been added, and can be imposed in tandem. The toolbox produces IRFs, FEVD, historical decompositions, and structural shocks, and can be easily modified to add new identifications.

MIDAS Nowcaster

A tool to nowcast quarterly data with monthly indicators, using a US consumption example. Pulls data directly from FRED from a list of codes — any target quarterly variable or monthly high-frequency indicator can be used. Calculates the optimal lag length to minimize out-of-sample RMSE for each indicator and whether the specification improves with an AR term included. Optimally combines the predicted nowcast using either MSE or RMSE weighting. Currently set up to nowcast US consumption using monthly PCE data, non-farm payrolls, and real retail sales.