Forecast Surveys

Download Report

Transcript Forecast Surveys

Forecast Surveys
Introduction
• Two Philadelphia Fed surveys of private-sector
forecasters
– Livingston Survey
– Survey of Professional Forecasters (SPF)
• Provide timely forecasts for policymakers and
economic analysts
• Used to answer numerous research questions
Introduction
• Philadelphia Fed economists pushed to have
surveys here
– Livingston Survey: Don Mullineaux (?)
– SPF: Dean Croushore
• Current research department keeps them here
& makes them ever more useful
– Loretta Mester (research director)
– Tom Stark (maestro)
Introduction
• Both surveys
– Available to the public at no charge (Fed’s public
education mission)
– Used by news media, policymakers, economic
analysts, labor unions, consumers
Introduction
• Livingston Survey
– Begun in 1946
– Joe Livingston, journalist
– 1970s: economists discover survey
– 1978: Philadelphia Fed takes over database
– 1990: Philadelphia Fed takes over administration
of survey
Introduction
• SPF
– ASA/NBER survey begun in 1968
– Victor Zarnowitz & other founders
– ASA/NBER folds survey in 1990
– Philadelphia Fed takes over in 1990
– Renamed Survey of Professional Forecasters
What Variables Do the Participants Forecast?
Both surveys:
•
•
•
•
•
Nominal GDP
Real GDP
Inflation (CPI)
Unemployment rate
Industrial production
• Interest rate: 3-month Tbills
• Interest rate: 10-year Tnotes
• Corporate profits after tax
• Housing starts
• Business fixed investment
What Variables Do the Participants Forecast?
Livingston survey:
•
•
•
•
•
•
Producer price index
S&P 500 stock prices
Average weekly earnings
Retail trade sales
Auto sales
Prime interest rate
• Average over the next 10
years:
– Real GDP growth
– CPI inflation rate
What Variables Do the Participants Forecast?
SPF:
•
•
•
•
GDP price index
Payroll employment
Interest rate: AAA bonds
Remaining GDP
components
• Inflation rates:
– CPI excluding food & energy
prices (core CPI)
– PCE price index
– PCE price index excluding
food & energy prices (core)
• Probability that real GDP
will decline
• Distribution forecasts:
–
–
–
–
Real GDP
GDP price index
Core CPI
Core PCE price index
• Long-term forecasts:
– CPI price index
– PCE price index
What Variables Do the Participants Forecast?
SPF:
• First Quarter:
– Long-term projections
•
•
•
•
•
Real GDP growth
Productivity growth
Stock returns
Long-term bond returns
Short-term bond returns
• Third Quarter:
– Estimates of natural rate
of unemployment
• Special Questions
– Declines in house prices
– Effects of stimulus
package
– Effect of Y2K
Anonymity
• Both surveys promise anonymity
– Can’t match forecaster to a forecast
– Allows forecasters to reveal their true forecast,
without attribution
– Prevents herding and extreme forecasts
Evaluating the Survey Forecasts
• Are the forecasts accurate?
• Statistical tests
– Unbiasedness (forecast errors have zero mean)
– Efficiency (forecast errors are uncorrelated with
information known when forecast was made)
Evaluating the Survey Forecasts
• Unbiasedness
– Over long samples, forecasts of major variables
appear unbiased (Figure 1)
Evaluating the Survey Forecasts
• Unbiasedness
– Over long samples, forecasts of major variables
appear unbiased (Figure 1)
– Formal statistical tests:
f
t
t
t
Y    Y  
• Test: α = 0, β = 1
Evaluating the Survey Forecasts
• Unbiasedness
– Or:
Yt  Yt     t
• Test: α = 0
f
Evaluating the Survey Forecasts
• Sub-periods show poor performance
Evaluating the Survey Forecasts
• Sub-periods show poor performance
• Persistent forecast errors in 1970s and 1990s
• Those periods require explanation: failure to
recognize impact of faster money growth;
errors in evaluating natural rate of
unemployment and rate of potential GDP
growth
Evaluating the Survey Forecasts
• Inefficiency with respect to other variables
• Test: forecast errors should be uncorrelated
with other variables known when survey was
conducted
– Example from Ball-Croushore (Figure 3)
Evaluating the Survey Forecasts
• Inefficiency with respect to other variables
– Example from Ball-Croushore (Figure 3)
– SPF forecasters seem not to change output
forecasts enough in response to change in
monetary policy
Evaluating the Survey Forecasts
• Accuracy of probability distribution forecasts
– Results from Diebold-Tay-Wallis (1999)
• Forecasts are reasonably accurate
• But too much probability of a large reduction in
inflation
• Persistent inflation forecast errors
Evaluating the Survey Forecasts
• Conclusion
– SPF & Livingston survey forecasts fairly accurate—
passing most tests
– Some imperfections