Geoscience Reference
In-Depth Information
while employment in nonbasic industries, such as retail trade and state and local
government, is predicted using endogenous variables (e.g., Puget Sound personal
income and the unemployment rate). In addition to employment, the regional model
forecasts the unemployment rate, personal income, wages and salaries, retail sales,
population, the consumer price index, housing permits, home prices, and the
apartment vacancy rate. The predictions of the exogenous variables come from
the Blue Chip consensus forecast for the U.S. economy (
Moore, monthly
).
Over the past 20 years, the one-year-ahead absolute prediction error for employ-
ment has averaged 0.9 %. This compares favorably with the 1.2 % average error for
real U.S. Gross Domestic Product (GDP) associated with the Blue Chip forecasts
over the same period. These two variables are equally difficult to predict given that
the ranges in their yearly growth rates since 1990 are similar. While real GDP
growth rates have varied from
2.8 % in 2009 to +4.8 % in 1999, regional
employment growth rates have varied from
4.0 % in 2009 to +5.1 % in 1997.
This does not mean that the regional employment prediction errors are all tightly
clustered around 0.9 %. There are two errors—a 2.7 % error in 2001 and a 3.2 %
error in 2009—that are significantly larger. Both errors are associated with the onset
of a recession, reflecting another fact of life in forecasting: national and regional
economists have difficulty predicting downturns.
Ironically, the large forecasting errors caused by the Great Recession provide an
opportunity to illustrate the openness of the Puget Sound economy. Consider the
following question: under what conditions would we have predicted the regional
recession? A conditional prediction test provides an answer.
In an ex ante forecasting test conducted at the end of 2009 (Conway
2009
), it was
assumed that back in December 2005 there was perfect knowledge about the course
of the U.S. economy, Boeing, and Microsoft—the exogenous variables in the
regional forecasting model—through 2009. With the benefit of known exogenous
variables, the model calibrated in December 2005 would have predicted the regional
recession with relatively small errors (Table
8.7
). At the end of the 4-year forecast
period, for example, the conditional accumulated prediction error for employment
was
0.6 %, much less in absolute terms than the 6.2 % actual accumulated error.
The conditional accumulated prediction errors for personal income and population
were
2.8 % and 0.7 %, respectively, both small miscues for a 4-year projection.
11
A perhaps counterintuitive outcome of the Great Recession was its negligible
impact on Puget Sound population growth. In September 2007, prior to the down-
turn, it was predicted that population would increase by 254,000 over the next
11
During the Great Recession, Washington state government produced economic and tax revenue
forecasts that not only exhibited extremely large errors but were also unreasonable. A state version
of the Puget Sound Forecasting Model (Conway
2010
) was built to provide alternative projections
through FY 2013. The alternative predictions turned out to be significantly more accurate. A
contributing reason for the relative precision of the alternative forecasts was the fact that the Blue
Chip panel of national economists, which called for a slow national recovery between 2010 and
2013, essentially hit the mark.
