1www.bea.gov

Report on the Income- and Product-Side

Estimates of Output Growth

Comments

Steve Landefeld

Brookings Panel on Economic ActivityMarch 19, 2010

2www.bea.gov

Report on the Income- and Product-Side Estimates of Output Growth

▪ Excellent and imaginative paper. BEA appreciates work such as this on measurement issues relating to its accounts. External research, complemented by

research at BEA, has been the source of a wide range of statistical improvements ranging from chain to hedonic indexes.

bea.gov 3

GDP and GDI Source Data

Both income and expenditures have

measurement strengths and weaknesses BEA devotes substantial resources to updating and

maintaining both measures

Nevertheless, we think there are good reasons for us to

feature a single measure GDP measure and for

preferring the expenditure-side measure.

In general, we feature the expenditure estimates

because of the timeliness of the quarterly source

data and the consistency of that data with BEA

concepts and definitions and with the annual and

benchmark data

bea.gov 4

GDP and GDI Source Data: Timeliness

Quarterly source data for GDP is more timely

than GDI Improvements in coverage being incorporated in GDP

(services)

Improvements in coverage being incorporated in GDI

(wages)

Picture is more mixed for annual and

benchmark data

QCEW quarterly employment and unemployment tax

data for GDI available with one quarter lag

Census annual business statistics data available with

1-2 year lag

IRS annual data are available with a 2-3 year lag

bea.gov 5

GDP vs. GDI – Source Data for the Third Estimate of the Quarter

GDP23% based on

judgmental trend

77% based on early source data

|||||||||||||||||||||||||

63% based on judgmental trend

37% based on early source data

GDI

bea.gov 6

GDI Source Data

Judgmental trend

Early source data Table 1. GDI source data – estimates for 2007

(percent of GDI)

Description of early source data / estimation method

Gross domestic income Compensation of employees Wages and salaries Nonsupervisory & production workers 17.1% BLS Current Employment Statistics: payroll survey Supervisory/nonproduction workers 20.6% Judgmental extrapolation based on payroll survey Government 7.7% Payroll survey employment and ECI Supplements 10.3% Judgmental trend extrapolation Taxes on production and imports, less subsidies Property taxes 2.8% Judgmental trend extrapolation Other 4.1% Federal Monthly Treasury data; Census data for sales taxes Net interest and misc. payments 6.8%

FDIC data for commercial banks; judgmental trend extrapolation based on interest rates for most of the remainder

Business current transfer payments 0.7% Judgmental trend extrapolation Proprietors' income 7.8%

Judgmental trend extrapolation based on BLS payroll survey, Census data, and other indicators

Rental income of persons 1.0% Mixture of actual source data and judgmental extrapolation Corporate profits 8.5% Census Quarterly Financial Report, FDIC, Compustat data Current surplus of government enterprises 0.0% Judgmental trend extrapolation Consumption of fixed capital 12.5% Judgmental extrapolation based on BEA capital stocks Total 62.6% 37.4% Percent based on early source data that are conceptually consistent with annual/benchmark data 11.8%

bea.gov 7

GDP and GDI Source Data: Consistency

The source data for GDP is mainly data Census collects using consistent definitions and surveys that reflect BEA’s needs and that can be consistently benchmarked to the once-every-five years Economic Census.

The source data for GDI is mix of mainly administrative data that are collected for non-statistical purposes using multiple definitions across the sources and over time.

There may be significant timing differences between

the source data for expenditures and income.

bea.gov 8

Growth in Wages and Salaries-- Census, BLS, and SOI

-2

0

2

4

6

8

10

1999 2000 2001 2002 2003 2004 2005 2006 2007

Year

Per

cen

tag

e

Census Business Patterns (CBP) BLS QCEW SOI Salaries and Wages

Different Measures of Wages and Salaries

bea.gov 9

Different Measures of Profits

Percent Change of Profits Measures

-100%

-80%

-60%

-40%

-20%

0%

20%

40%

60%

1999 2000 2001 2002 2003 2004 2005 2006 2007

Year

Y/Y

Per

cen

t C

han

ge

S&P Operating Earnings NIPA Profits (current production)

SOI (Total Receipts less Total Deductions) QFR Profits (Mining,MFG,Trade)

bea.gov 10

GDP and GDI Source Data: “Accuracy”

The Census-based GDP estimates are within 1

percent of the comprehensive revisions based on

the once-every-five year Economic Census.

The GDI estimates are also within 1 percent of the

comprehensive revisions based on the once-every-

five year Economic Census. Despite being estimated using very different source

data and methods the closeness of these two 5-year

extrapolations suggest that for trend growth, the

two estimates appear to tell a consistent story.

bea.gov 11

GDP vs. GDI Tests

Cyclical indicators used may not be a good measure of output turning points: NBER Cycles: Use and reliance on GDI data

Employment and real personal income less transfers as a monthly measure of output

S&P 500: Good financial indicator, but checkered history as cyclical indicator

May move with GDI, if capital gains and losses are “leaking” into GDI

This may help explain relation of the Statistical Discrepancy to the business cycle

Employment and Unemployment: Employment determinant of wages and salaries in GDI

bea.gov 12

GDP vs. GDI Tests

Historically revisions to GDI may have helped reduce revisions to GDP, but gain would have been small Use of GDI could have reduced early GDP estimate

MAE by 0.07 percentage points, from 1.25 to 1.18, 1994-2006

Revision pattern over this period reflected many changes in source data and methods. With major new source data and methods being

introduced in GDP (all services) and GDI (all employee wages), it is difficult to predict result of using averaging going forward.

bea.gov 13

GDP vs. GDI Tests

Okun’s law isn’t a law but an empirical relationship. Relationship has varied over time and over the business cycle;

progressive reduction in the responsiveness of unemployment

to GDP

Okun’s law seems to be broken for both GDP and GDI GDI flat to falling mid 2006 to mid 2008, but unemployment

falling

GDP slower growth but rising 2006 to mid 2008 and

unemployment falling

Both GDP and GDI steep fall since mid 2008 and steep rise in

unemployment.

Unlikely that revision to GDP would completely erase

productivity gains and the good economic explanations for

the changes occurring in labor markets.

bea.gov 14

GDP vs. GDI Tests

GDP and GDI early estimates produce the same general picture of economic activity. GDP available 1-2 months before first GDI estimate.

The last two cycles, GDP and GDI latest estimates present the same general picture, but with significant differences in specific quarters. Most recent period: Slowing of growth in 2007 both

GDI and GDP Peak to Trough cumulative decline, GDP -3.1%; GDI -4.2%. Significant revisions to come:

GDI: New annual IRS data for 2008 and revised data for 2007

GDP: New annual Census data for 2009 and 2008 and revised data for 2008

bea.gov 15

Early Real GDP and GDI Estimates (3rd Estimates)

-8

-6

-4

-2

0

2

4

6

8

10

1984

Q1

1984

Q4

1985

Q3

1986

Q2

1987

Q1

1987

Q4

1988

Q3

1989

Q2

1990

Q1

1990

Q4

1991

Q3

1992

Q2

1993

Q1

1993

Q4

1994

Q3

1995

Q2

1996

Q1

1996

Q4

1997

Q3

1998

Q2

1999

Q1

1999

Q4

2000

Q3

2001

Q2

2002

Q1

2002

Q4

2003

Q3

2004

Q2

2005

Q1

2005

Q4

2006

Q3

2007

Q2

2008

Q1

2008

Q4

2009

Q3

Pe

rce

nt

GDI GDP

bea.gov 16

Latest Real GDP and GDI Estimates

bea.gov 17

Early Real GDP and GDI Estimates (3rd estimates)

bea.gov 18

Latest Real GDP and GDI Estimates

bea.gov 19

Early GDP and GDI Estimates(3rd estimates)

Third Estimates of Real GDP and Real GDI: Current Recession

-8

-6

-4

-2

0

2

4

6

8

2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4

%∆GDI_R$(3rd) %∆GDP_R(3rd)

bea.gov 20

Latest Real GDP and GDI Estimates

Latest Estimates of Real GDI and Real GDP: Current Recession

-8

-6

-4

-2

0

2

4

6

8

2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4

%∆GDI_R$(latest) %∆GDP_R(latest)

bea.gov 21

Averaging GDP and GDI

BEA will incorporate a better means of presenting data on GDI, without unduly confusing the general community of users.

We aren’t convinced that averaging the estimates would provide enough benefits to our broad range of users to make it worth the costs in terms of larger revisions to the early estimates, explaining the approach to users, etc.

bea.gov 22

Next Steps

Incorporation of the new Census data on services and more experience with the new BLS payroll and QCEW data should produce significant improvements in GDP and GDI.

Further research into balancing, now being conducted in the industry accounts.

Research on finance, and the role of capital gains, could prove helpful in better understanding and addressing the statistical discrepancy.

Data Synchronization could prove extremely helpful in resolving inconsistencies between BLS and Census data.