View original document

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

short essays and reports on the economic issues of the day
2009 ■ Number 9

How Accurate Are Forecasts in a Recession?
Michael W. McCracken, Economist
ndividual households and firms, as well as local, state,
and federal governments, make economic decisions
based on their view of the future. For a household,
this may entail deciding whether to apply for a mortgage
based on the expectation of having a job in the future. For
a firm, a decision to invest in the building of a new plant
rests on expected future demand for its products. Similarly,
federal agencies make decisions about developing infrastructure based on expectations of future revenue.


“On average, professional forecasters
have tended to be less accurate when
the U.S. economy was in recession.”
Firms and government agencies (and, to a lesser extent,
households) look to professional forecasters for advice on
the future state of the economy. This essay looks at how a
recessionary environment affects the accuracy of economic
forecasts, which in turn affects economic decisionmaking.
Although there are numerous professional economic
forecasting agencies, the Survey of Professional Forecasters

(SPF) is a publicly available forecasting source that is commonly used and discussed in various media outlets. This
discussion uses the mean of those forecasts as a proxy for
professional forecasters in general.1 In total, we review 26
years’ worth of quarterly, 1-year-ahead mean SPF forecasts
from 1981:Q3 through 2007:Q3. The subsequent forecast
errors—ranging from 1982:Q3 through 2008:Q3—are constructed using the most recent and updated data. We use
the squares of these errors as a measure of their accuracy.
The chart shows the squared forecast errors associated
with the mean SPF forecast of U.S. gross domestic product
(GDP) growth2 and the unemployment rate. The shaded
areas indicate U.S. recessions, as determined by the National
Bureau of Economic Research. Over the time frame shown
in the chart, on average, professional forecasters have tended
to be less accurate when the U.S. economy was in recession.
During a recession, the mean squared errors (MSEs)
associated with forecasts of GDP growth and the unemployment rate are 0.58 and 1.95, respectively. In each case,
the MSEs are four times larger than those made when the
economy is not in a recession. Even if the very large errors
during and immediately after the 1981 recession are omitted

Squared Forecast Errors


Unemployment (left axis)
GDP (right axis)















Economic SYNOPSES

Federal Reserve Bank of St. Louis

from the calculations, the MSEs are still roughly twice as
large—a smaller but still significant difference.
One main contributor to the deterioration in the forecasts is an inability to detect turning points—that period
when the economy shifts from being in a recession to not
being in a recession—and vice versa. Another contributor
is a clear bias in the forecasts during a recession. Forecasts
of GDP growth and the unemployment rate both generally
tend to be overly optimistic: Forecasts of GDP growth tend
to be too high, whereas those for the unemployment rate


tend to be too low. Perhaps not surprisingly, the degree of
accuracy of the two series tends to move together; the correlation between the two series is roughly 80 percent. The
unfortunate conclusion is that during a recession, economic
decisions are not only more important, but also more difficult to make. ■

The SPF is available at
2 The SPF predicts gross national product (GNP) through 1991 and GDP thereafter.

Posted on February 6, 2009
Views expressed do not necessarily reflect official positions of the Federal Reserve System.