View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.


, May 8, 2020 - Federal Reserve Bank of Richmond

Economic Impact of COVID-19
May 8, 2020

Forecasting the COVID-19 Epidemic for the U.S.
Article by: Paul Ho, Thomas A. Lubik and Christian Matthes

Download article

Updated Estimates and Forecasts
June 10, 2020: We have updated estimates and forecasts of infection and mortality rates
in the United States and in the constituent states of the Fifth District. The current update is
based on data up to June 6 and now also includes mortality projections for the latter. We
reestimate the model with the recent data so that the projections reflect both the influence
of revised model estimates and more data. A full set of graphs for U.S. and Fifth District
forecasts is available online.
For the entire United States, we now project median fatalities of 184,000 with a 95 percent
range of 175,000 to 193,000 by early September. The corresponding number of total
infections is 3.1 million. Even three months out, we expect the United States to have
10,000 new cases and almost 600 deaths daily. Our forecast intervals have shifted slightly
upwards, possibly on account of relaxed social distancing measures in many parts of the
United States since early May. Otherwise, the model performs exceedingly well in that
projections for daily infections and fatalities become more precise and data realizations
continue to lie within error bands.
Over the same horizon, we project 3,400 cumulative fatalities in Virginia, 2,800 in North
Carolina, 5,500 in Maryland, and 1,200 in South Carolina, with the District of Columbia
and West Virginia below the 1,000 mark, although the per capita rates in the former are
exceedingly high. Although forecast intervals have tightened, the data flow for North and
South Carolina has started to fall outside of the error bands from the previous forecast,
which led to a considerable revision in the estimates. The pattern of new infections and
deaths for these two states suggests the effect of early relaxation of lockdown measures,
although the per capita rates are still comparatively low. We are developing a richer model
for all states that allows forecasts to adjust to variables including the amount of social
distancing. Preliminary results and further discussion are here.