Transcript Document

GDP measurement issues
Graeme Walker
Head of National Accounts
Productivity Puzzle Seminar: 16 October 2012
Outline
• Why is GDP revised?
• How is it compiled?
• How does this lead to revisions?
• When might we expect revisions?
• Size of GDP revisions
• Reference to the Productivity Puzzle
• Impact of Revisions
• Conclusions
• Measurement issues
• What’s next?
How GDP is compiled (1)
• Three approaches
• Output, Expenditure and Income
• Perfect statistical world – all equal
• Many sources
• Monthly, Quarterly, Annual
• Various timings
• Monthly survey (IoP around 6 weeks)
• 44,000 questionnaires
• Annual tax returns (15 months)
How GDP is compiled (2)
• Preliminary estimate (25 days)
• Based exclusively on output, 44% data
• Second estimate (8 weeks)
• Growth based on output, 83% data
• Quarterly National Accounts (13
weeks)
• Based on output, 92% data
• Revisions to previous periods
How GDP is compiled (3)
• Balanced supply and use
• Framework for confronting differences
• Levels not growths
• First balance after around 18 months
• Blue Book 2012 balanced 2010
• Second balance is usually based on
‘final’ data except for methodological
changes
Why is GDP revised?
•
•
•
•
•
•
Output source data revised
Revised seasonal factors
Annual chain-linking changes weights
Supply and use balancing
New methods
New international frameworks
When might we see revisions?
• For 2010 and earlier, nothing before
June 2013
• Supply and Use balancing
• Some methodological changes
• For 2011 and 2012, the Quarterly
National Accounts releases in Dec
2012 and March 2013
• New data but still based on output
• 2011 balanced for the first time in Jun
2013
Size of GDP Revisions
• Article published today
• Update of comprehensive analysis published in
December 2009
• Normally, revisions are compared over 5 years
• But look at 2 years, to cover downturn
• First estimate of GDP still unbiased and
not significant
• Signs that revisions are getting a bit larger
• Subsequent revisions are largely
methodological and cannot be predicted
• Size of possible future revisions
GDP Revisions
Mean revisions between T and T + 24 months
0.20
Mean revisions - Per cent growth
0.15
0.10
0.05
0.00
-0.05
1961Q2 to 1971Q4
1972Q1 to 1983Q4
1984Q1 to 1994Q4
1995Q1 to 2004Q4
2005Q1 to 2010Q2
GDP Revisions – Ignoring sign
Mean absolute revisions between T and T + 24 months
0.9
0.8
0.7
Mean revisions - Per cent growth
0.6
0.5
0.4
0.3
0.2
0.1
0
1961Q2 to 1971Q4
1972Q1 to 1983Q4
1984Q1 to 1994Q4
1995Q1 to 2004Q4
2005Q1 to 2010Q2
GDP level
Index, 2008 Q1 = 100
101
100
99
98
97
96
95
BB12
94
BB11
BB10
93
BB09
Q2
2012 Q1
Q4
Q3
Q2
2011 Q1
Q4
Q3
Q2
2010 Q1
Q4
Q3
Q2
2009 Q1
Q4
Q3
Q2
2008 Q1
92
GDP Level – Future revisions?
Index, 2008 Q1 = 100
101
100
99
98
97
96
Illustrative future
revisions
BB12
95
94
BB11
93
BB10
Q2
2012 Q1
Q4
Q2
2011 Q1
Q4
Q3
Q2
2010 Q1
Q4
Q3
Q2
2009 Q1
Q4
Q3
Q2
2008 Q1
Q3
BB09
92
Conclusions
• GDP revisions remain unbiased and not
significant
• But some signs they are becoming larger
and so, monitoring will continue
• Even so, unlikely that future revisions to
GDP could explain more than one
percentage point of the productivity puzzle
at most
Measurement issues for initial
estimates of GDP
• Why might revisions be getting larger?
• More difficult at ‘turning points’ or when growth is more variable
• Estimate based on deflated turnover
• No short term estimate of intermediate consumption
• Assumption that it moves in line with turnover
• Deflators?
• Difficulty of measuring small businesses
• Register of businesses based on VAT registrations and
employer tax records
• Under-coverage adjustment for self employed not on business
register
• Difficulty measuring some sectors
• Financial services
• Construction
What’s next?
• BB 2013 – More methods improvements; full
scope to be published later this year
• BB 2014 – New National Accounts framework
(ESA10)
• Analyse revisions by reason rather than over
time (feasibility)
• Continue to monitor revisions
• GDP improvement plan
• Part of continuous improvement
Questions
Email: [email protected]