RealTime Data at the Federal Reserve Bank of Philadelphia

1
Real-Time Data at the Federal Reserve Bank of
Philadelphia

• 2008 World Congress on National Accounts and
  Economic Performance Measures for Nations
• Washington, DC
• May 13 17, 2008
• Tom Stark
• Federal Reserve Bank of Philadelphia

2
Real Time?

• History Observations on economic variables,
  measured prior to revision
• Philadelphia Fed real-time data set
• Forecasts Projections based on real-time history
• Philadelphia Fed forecast surveys

3
Overview of Talk

• Real-Time Data Sets at the Philadelphia FRB
• Forecast Surveys
• Realizations

4
Overview of Talk

• Real-Time Data Sets at the Philadelphia FRB
• Forecast Surveys
• Realizations
• Uses of Real-Time Data at the Philadelphia FRB
• Analyzing revisions
• Evaluating forecast surveys
• Testing the forecast implications of
  macroeconomic theories
• Carrying out historical policy analysis

5
Overview of Talk

• Real-Time Data Sets at the Philadelphia FRB
• Forecast Surveys
• Realizations
• Uses of Real-Time Data at the Philadelphia FRB
• Analyzing revisions
• Evaluating forecast surveys
• Testing the forecast implications of
  macroeconomic theories
• Carrying out historical policy analysis
• Some Examples
• Revisions to the U.S. personal saving rate
• Forecast accuracy in the Survey of Professional
  Forecasters
• Using the PIH to forecast income in real time
• FOMCs response to inflation expectations

6
Real-Time Forecasts

• Livingston Survey
• Semiannual June 1946 to present
• Focus on business indicators - Most popular CPI
• Survey of Professional Forecasters
• Quarterly 1968 Q4 to present
• Focus on U.S. national accounts Product side
• Greenbook Forecasts (Federal Reserve Board)
• Prior to each FOMC meeting, 1966 - present
• PDF files provide wide range of variables
  forecast

7
Real-Time History Philadelphia Fed Real-Time
Data Set for Macroeconomists

• Started in early 1990s as a research project with
  Dean Croushore
• Provides snapshots of the data before revision
• Has since become an ongoing Bank project at the
  Philadelphia FRB

8
Real-Time History (cont.)Variables Included at
the Phila. FRB

• NIPA Income and Product Side
• Monetary Aggregates
• Business Indicators (IP, Capacity Utilization,
  Housing Starts)
• Labor Market (Unemployment, Employment, Hours
  Worked)
• Aggregate Price Indexes (GDP, CPI, PCE)

9
Uses of Real-Time Data at the Philadelphia FRB

• Analyzing Revisions
• Evaluating Survey Forecasts
• Testing Theories
• Conducting Historical Policy Analysis

10
(1) RevisionsU.S. Personal Saving Rate

• Nakamura and Stark (February 2007) Philadelphia
  Fed Working Paper No. 07-8
• Key Findings on Personal Saving Rate
• Early estimates very unreliable
• Large upward revisions in benchmarks

11
U.S. Measured Personal Saving Rate ()Large
Upward Revisions1965 to 2005
2005 Q3 Vintage
Advance Estimates
12
U.S. Measured Personal Saving RateDistribution
of RevisionsLatest Available Minus Advance
Estimate1965 to 2005
Mean 2.44 Std Dev 1.88 MSE 9.52
13
U.S. Measured Personal Saving RateDistribution
of RevisionsLast-before-Benchmark Minus Advance
Estimate1965 to 2005
Mean 0.08 Std Dev 1.06 MSE 1.11
14
(2) Evaluation of Survey ForecastsSurvey of
Professional Forecasters

• Two Key Issues
• Revised or unrevised realizations
• Does it matter?
• If so, how much?
• Information sets for estimating and forecasting
  benchmark models

15
Revised or Unrevised Realizations?Nowcast RMSE
Q/Q Real GDP Growth (Annual Rate, PPs.)Survey
of Professional Forecasters, 1985-2005
16
Revised or Unrevised Realizations?Four-Quarter-Ah
ead RMSE Q/Q Real GDP Growth (Annual Rate,
PPs.)Survey of Professional Forecasters,
1985-2005
17
Revised or Unrevised Realizations?Nowcast
RMSEQ/Q GDP Inflation (Annual Rate, PPs.)Survey
of Professional Forecasters, 1985-2005
18
Revised or Unrevised Realizations?Four-Quarter-Ah
ead RMSEQ/Q GDP Inflation (Annual Rate,
PPs.)Survey of Professional Forecasters,
1985-2005
19
Benchmark ComparisonsTiming of Information Sets

• First Quarter Survey
• Forecasts for Q1,,Q4,Q1

January
February
March
NIPA Release for Q4 Questionnaire sent R-Time
Data Set for NIPA
Survey deadline
Other January data Retail Sales, CPI,
PPI Housing Starts Industrial Production
January data Labor market Interest rates
20
Benchmark Comparisons AR ModelsRMSE (SPF / AR),
1985 2005Q/Q Real GDP Growth
21
Does The Relative Value of Nowcast Information
Decline as Data Are Revised?RMSE (SPF / AR),
1985 2005Q/Q Real GDP Growth
22
(3) Testing Economic TheoriesPermanent Income
Hypothesis (PIH)

• Nakamura and Stark (2007) Philadelphia Fed
  Working Paper 07-8
• Test Irelands Model of Forecasting with the PIH
  in Real Time
• Key Findings
• Model works well when estimated on revised data
• Not so good in real time

23
The Model and Methodology

• Estimate VAR in income growth and the saving rate
• Impose restrictions implied by PIH
• Run horse races between VAR-PIH and AR forecasts
  for income growth
• Using fully revised data
• Using data in hand in real time

24
Results Forecasts for Income GrowthRMSE
(PIH/AR) Ratio1971 to 1981
25
Results Forecasts for Income GrowthRMSE
(PIH/AR) Ratio1971 to 1981
26
(4) Historical Policy AnalysisHow Has the Fed
Responded to a Rise in Inflation Expectations?

• Leduc, Sill, and Stark (2007) Journal of
  Monetary Economics
• How has the FOMC responded to expected inflation
  before and after 1979?
• Requires real-time measure of expectations for
  inflation Livingston Survey
• Findings Strong response post-1979.

27
The Model and Methodology

• VAR in actual and expected CPI inflation,
  unemployment, short-term interest rate, and
  commodity prices
• Real-Time Data Expected inflation from
  Livingston Survey
• Examine impulse responses to a one-time shock to
  expected inflation
• Sample period cut at 1979

28
We Examine the Effect That a One Percentage Point
Shock to (Livingston) Inflation Expectations.
Effect of Shock on Expectations for Inflation
Post-1979
Pre-1979
29
Has on Inflation and the Real Rate
Pre-1979
Post-1979
Inflation
Real Rate
30
Concluding RemarksUpcoming Changes

• May 2, 2008 INDUSTRY dummy variable added to
  micro data set for SPF forecasters
• Allows analysts to cut the data according to
  panelists employed in financial and nonfinancial
  sectors
• Summer 2008 New data-download pages for SPF and
  real-time data sets
• Easier downloads for users
• Summer 2008 Forecast error statistics for
  (mostly) all variables in SPF
• Ongoing New variables and more frequent vintages
  in real-time data set joint with Dean Croushore
• Monthly vintages for NIPA variables
• Price indexes for PCE and core PCE
• Detailed components from BEAs personal income
  report

31

• Tom Stark
• Tom.Stark_at_phil.frb.org