Nowcasts for the U.S. output gap are calculated following the approach described in
Berger, Morley, and
Wong (2023, Journal of Econometrics) using a model that includes the U-2 unemployment rate
instead of the unemployment rate, as in Berger,
Boll, Morley, and
Wong (2022, Oxford Open Economics). Please cite these papers when referring to the
estimates reported here.
The latest nowcast reported above is displayed in the time series plot using an orange line extending from
the
realized estimates of the output gap based on data up to and including the latest quarter for which real GDP
has been released.
Conditional forecasts of future values of the output gap can be toggled on or off in the plot by clicking on
the label in the legend. These are displayed with a green line.
Exact values of estimates can be seen by hovering over or tapping on each point in the plot.
The data file linked to above contains three columns: (1) the quarter of an
estimate, (2) the output gap estimate (%), and (3) whether or not real GDP has been realized yet in a given
quarter.
The estimates are automatically updated daily and usually within 30 minutes of major data releases.
The date (UTC time zone) of the latest nowcast is reported directly below the time series plot.
Parameter estimates
are based on the full sample for which real GDP has been released, but not
including 2020Q1-2020Q3 in the forecast evaluation used for calculation of
the shrinkage hyperparameter in order to avoid any distortions from Covid-related outliers.
For details of the indicators used in calculating the output gap estimates or information on historical nowcasts
for the given quarter, click the
'Indicators' or 'Historical Nowcasts' dropdown, respectively.
The table below contains the most recent values of the indicators that feed into the output gap
estimates. Note that consumer sentiment may be a
preliminary measure and there can be revisions to IP growth, housing starts, and real GDP growth over time.
Also, we note that the interest rates used in calculating spreads are measured
only to two decimals.
Below are three CSV data files that can be placed in the same directory as the MATLAB code linked to at
the bottom of this webpage in order to replicate the nowcast estimates.
These data are mostly gathered from FRED, except for some indicators reported on FRED with a lag or
truncated timelines, which we scrape from other sources.
Contents by column (denoted by FRED codes - starting from 1967):
Monthly data: date, FEDFUNDS, UMCSENT, U2RATE, CPIAUCSL, INDPRO, HOUST, SP500
Monthly data for spreads: date, BAA, AAA, DGS10, DGS1
Quarterly data: date, GDPC1
The table below contains the historical nowcasts for the quarter currently being nowcast.
For previous days, the final nowcast of each day (based on the UTC time zone) is given in the table.
The most recent nowcast given in the table may be revised within the day.
To download the above historical nowcasts for the current quarter click the download button below, or to
download another quarter's historical nowcasts, change the selection from the dropdown.
Contents by column:
Date of when the nowcast estimate is made, nowcast (%)
Tino Berger, James Morley and Benjamin Wong, "Nowcasting the Output Gap," Journal of Econometrics,
vol. 232, issue 1, January 2023, pp. 18-34.
Link to journal
article (open access)
Tino Berger, Paul David Boll, James Morley and Benjamin Wong, "Cyclical
Signals from the Labor Market," Oxford Open Economics, vol. 1, 2022.
Link to journal article (open access)