Evaluating Consensus Forecasts

We find value in tracking consensus forecasts
for growth, in part because these projections
are a good proxy for investors’ views. In this
article we evaluate the past performance of
consensus GDP forecasts and show how to
apply some of the results.
In general, consensus forecasts are a useful
guide for growth over the next few quarters.
They easily outperform simple benchmarks,
implying that they contain useful information.
However, the consensus does not seem to be
able to systematically predict growth over
longer periods. In part this likely reflects the
greater probability of policy changes and other
shocks over these horizons.
A few themes stand out from 2011 forecasts.
First, growth views swung widely—there was
unusually large variation in GDP forecasts for
a non-recession year. Second, forecasters
sharply reduced their medium-term growth
views. Third, the forecast error for this year
(through Q3) was larger than average.
Because forecasts are sometimes off the mark,
it may be useful to characterize the uncertainty
around these projections. We describe a
straightforward method for creating a 70%
forecast confidence interval using historical
forecast errors (see chart at bottom right).
We also show how forecasters update their
views on new information. On average, about
40% of changes to forecasts for the current
quarter are passed through to the next quarter.
Despite a very busy data schedule, this week’s
releases had little impact on our expectations
for Q4 GDP growth; our “tracking estimate”
remains at 3.4%. The Federal Open Market
Committee’s statement this week upgraded its
assessment of the labor market slightly, but
remained cautious on activity overall.

来　　源：

高盛亚洲

撰写时间：

2011-12-16