Data Issues

1
Data Issues Revisions Norm
Williams December 12, 2003
2
U.S. Economic Data Revisions, What Are the
Issues?

• Businesses and policy makers rely on economic
  information and forecasts in making decisions to
  hire, invest, and provide or remove stimulus
• Trade-offs in collecting and providing
  data Timeliness versus Accuracy Cost versus
  Benefit
• Uncertainty in overall assessment of economys
  performance may lead businesses and policy makers
  to pull their punches

3
U.S. Economic Data Revisions, What Are the
Issues?

• Challenge to forecasters models specified to fit
  revised data may not work so well in real-time

• 2001 ExampleNot so much a case of Good Economy
  Goes Bad,but rather Bad Data Go Good

4
For much of 2001, many thought we would skirt a
recession.
Some press quotes during June July(names have
been withheld to protect the innocent)
and I do not think the United States economy is
in a recession but I think it is in a growth
slowdown. The economy is probably at or near
the bottom, and probably wont tip into
recession. The economy is not going into a
major recession
Source Bloomberg
5
Then in October many thought that 9/11 had tipped
us over into a recession.
Some press quotes during late October(names
have been withheld to protect the innocent)
9/11 will significantly harm the economy in the
third quarter and the fourth quarter and increase
significantly the likelihood that the economy is
in recession. Conference board survey 52 of
those polled said that the U.S. economy would
move into recession, up from 47 in September.
but, in late November, we were officially told
the recession began in March.
Source Bloomberg
6
Forecasting is made more difficult when the
current status of the economy is unclear.
Forecast and Actual Real GDP Growth,
Change from prior quarter at annual rate,
percentage points
3
Blue Chip mid-quarter forecast
2
1
0
-1
-2
-3
1Q2001
2Q2001
3Q2001
4Q2001
Source BEA, Blue Chip
7
Initial GDP reports have under-estimated growth
near the bottom of recessions, while showing
smaller misses near the peaks.
8
So How Can We Address These Issues?

• Knowledgeable users some evidence that private
  sector forecasters treat revisions as random,
  even though they may be somewhat predictable
• Trends in real-time information, like yield
  spreads and stock prices, do provide (slight)
  insight into direction of ultimate GDP revisions
• Fed relies significantly on supplemental
  anecdotes to inform its policy decisions (Beige
  Book, etc.)

9
So How Can We Address These Issues?

• BEA and BLS already taking major steps to improve
  quality (1) BEA adoption of chain weights in
  1996 comprehensive revision (2) BEA published
  12th comprehensive GDP revision this week (3)
  BLS major update to employment (nonfarm
  payroll) methodology in 2003
• NBER now using MAs monthly GDP estimates
• Would more funding for better statistics pay off?

10
By one estimate, paying for the 3 major US
statistical agencies already provides a
significant economic return, with plenty of room
to fund data quality improvements.