Transcript Discount Window Operations and the Implementation of

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
FEDERAL RESERVE BANK
OF PHILADELPHIA
Real Time?
• History: Observations on economic
variables, measured prior to revision
– Philadelphia Fed real-time data set
• Forecasts: Projections based on realtime history
– Philadelphia Fed forecast surveys
Overview of Talk
• Real-Time Data Sets at the Philadelphia FRB
– Forecast Surveys
– Realizations
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
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
– FOMC’s response to inflation expectations
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
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
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)
Uses of Real-Time Data at the
Philadelphia FRB
•
Analyzing Revisions
•
Evaluating Survey Forecasts
•
Testing Theories
•
Conducting Historical Policy Analysis
(1) Revisions:
U.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
U.S. Measured Personal Saving Rate (%)
Large Upward Revisions
1965 to 2005
14
12
10
2005 Q3 Vintage
8
6
4
2
Advance Estimates
0
-2
1965:03
1975:03
1985:03
1995:03
U.S. Measured Personal Saving Rate:
Distribution of Revisions
Latest Available Minus Advance Estimate
1965 to 2005
40
Mean:
35
Std Dev: 1.88
30
Count
2.44
MSE:
25
9.52
20
15
10
5
0
-4
-3
-2
-1
0
1
2
3
4
Revision (pct points)
5
6
7
8
U.S. Measured Personal Saving Rate:
Distribution of Revisions
Last-before-Benchmark Minus Advance Estimate
1965 to 2005
35
Mean:
30
Std Dev: 1.06
25
Count
0.08
MSE:
1.11
20
15
10
5
0
3
6
9
2
5
8
1
4
7
0
3
6
9
0.
0.
0.
0.
1.
1.
1.
2.
2.
2.
3.
3.
3.
3.
.3
-0
.6
.9
-0
-0
.2
-1
.5
-1
.8
-1
.1
-2
.4
-2
.7
-2
.0
-3
-3
.3
0
Revision (pct points)
(2) Evaluation of Survey Forecasts:
Survey 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
Revised or Unrevised Realizations?
Nowcast RMSE
Q/Q Real GDP Growth (Annual Rate, PPs.)
Survey of Professional Forecasters, 1985-2005
2.2
2
1.8
1.6
1.4
1.2
Initial
1Qtr Later
1Yr Later
Measure of Realization
2 Yrs Later
Today
Revised or Unrevised Realizations?
Four-Quarter-Ahead RMSE
Q/Q Real GDP Growth (Annual Rate, PPs.)
Survey of Professional Forecasters, 1985-2005
2.2
2
1.8
1.6
1.4
1.2
Initial
1Qtr Later
1Yr Later
Measure of Realization
2 Yrs Later
Today
Revised or Unrevised Realizations?
Nowcast RMSE
Q/Q GDP Inflation (Annual Rate, PPs.)
Survey of Professional Forecasters, 1985-2005
1.3
1.2
1.1
1
0.9
0.8
0.7
0.6
0.5
Initial
1Qtr Later
1Yr Later
Measure of Realization
2 Yrs Later
Today
Revised or Unrevised Realizations?
Four-Quarter-Ahead RMSE
Q/Q GDP Inflation (Annual Rate, PPs.)
Survey of Professional Forecasters, 1985-2005
1.3
1.2
1.1
1
0.9
0.8
0.7
0.6
0.5
Initial
1Qtr Later
1Yr Later
Measure of Realization
2 Yrs Later
Today
Benchmark Comparisons:
Timing of Information Sets
First Quarter Survey
Forecasts for Q1,…,Q4,Q1
January
February
March
NIPA Release for Q4
Questionnaire sent
Survey deadline
R-Time Data Set for
NIPA
Other January data:
Retail Sales, CPI, PPI
January data:
Housing Starts
Labor market
Industrial Production
Interest rates
Benchmark Comparisons: AR Models
RMSE (SPF / AR), 1985 – 2005
Q/Q Real GDP Growth
Measure of the Realization
Nowcast
4-Qtr
Ahead
Initial
1-Qtr
Later
1-Year
Later
2-Years
Later
Today
0.82
0.89
0.89
0.87
0.91
0.99
0.97
0.96
0.98
1.01
Does The Relative Value of “Nowcast Information”
Decline as Data Are Revised?
RMSE (SPF / AR), 1985 – 2005
Q/Q Real GDP Growth
Measure of the Realization
Nowcast
4-Qtr
Ahead
Initial
1-Qtr
Later
1-Year
Later
2-Years
Later
Today
0.82
0.89
0.89
0.87
0.91
0.99
0.97
0.96
0.98
1.01
(3) Testing Economic Theories:
Permanent Income Hypothesis (PIH)
• Nakamura and Stark (2007):
Philadelphia Fed Working Paper 07-8
• Test Ireland’s 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
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
Results: Forecasts for Income Growth
RMSE (PIH/AR) Ratio
1971 to 1981
Revised Data
1-Step
0.84
2-Step
0.85
4-Step
0.84
8-Step
0.91
Results: Forecasts for Income Growth
RMSE (PIH/AR) Ratio
1971 to 1981
Revised Data
Real-Time Data
1-Step
0.84
0.94
2-Step
0.85
0.99
4-Step
0.84
1.02
8-Step
0.91
1.06
(4) Historical Policy Analysis:
How 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.
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
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
1.5
1.5
1
1
0.5
0.5
0
0
-0.5
-0.5
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
…Has on Inflation and the Real Rate
Pre-1979
Post-1979
2
2
1.5
Inflation
1
1
0.5
0
0
-0.5
-1
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
8
9
10
11
12
8
9
10
11
12
2
1
1.5
0.5
Real Rate
1
0.5
0
0
-0.5
-0.5
-1
-1
1
2
3
4
5
6
7
8
9
10
11
12
1
2
3
4
5
6
7
Concluding Remarks:
Upcoming 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 BEA’s personal income report
Tom Stark
[email protected]
FEDERAL RESERVE BANK
OF PHILADELPHIA