Transcript Timeliness and average of revisions

Revisions in euro area GDP and retail trade:
Reliability, timeliness and infra-annual distribution
Heinz Dieden
European Central Bank – DG Statistics
OECD Short-Term Economic Statistics Working Party
(STESWP); Paris, 23-24 June 2008
Introduction
• Key elements
– Reliability/revisions
– Timeliness
– Infra-annual distribution of revisions
• Indicators
– GDP volume growth (q-o-q); sa & wda
– Constant retail trade turnover (m-o-m); sa & wda
• Euro area aggregates
– GDP: 2000 Q1 until 2006 Q4
– Retail trade 2000 January until 2007 April
1
Dataset
• Real-time database
– ECB’s revision database: from June 2001 onwards
– OECD’s MEI Release and Revisions database
– ECB’s Monthly Bulletin
• Timeliness
– Calendar days
– 30 days per month
2
Revision indicators
• Size of revisions
– Average absolute revisions
– Range of revisions
• Direction/sign of revision
– Average revision
– Changes in sign
– Upward/downward revisions
• Volatility of revisions
– Standard deviation
3
4
GDP
• Flash/first estimate versus second estimate
– Flash estimate from 77 days to 44 days
• Flash/first estimate versus annual update
– Annual update = estimate available in the 3rd quarter
of the following year
• Flash/first estimate versus current estimate
– Not full cycle of revisions for more recent data
Revisions for GDP
Table 1 Revisions in euro area GDP volume growth (2000Q1 – 2006Q4)
(percentages, if not otherwise indicated)
First estimate
First estimate
First value versus
versus second
versus annual
current estimate
estimate
update
*)
Size of revisions
Average absolute revision
0.06
0.09
0.12
Range of revision
[-0.1 , 0.2]
[-0.2 , 0.2]
[-0.2 , 0.5]
Sign of revision
Average revision
0.01
0.00
0.06
Changes in sign
0 out of 28
0 out of 24
1 out of 28
Upward/downward revisions
(number of cases)
16/12
11/13
17/11
Volatility of revision
Dispersion (standard deviation)
0.08
0.15
0.15
5
6
Evolution of GDP flash estimates
1.4
1.2
Range of estimates
Current estimate
First (flash) estimate
Annual update
1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
00Q1
02Q1
04Q1
06Q1
7
Timeliness and average of revisions
(GDP flash vs current estimate)
LHS: Timeliness (number of calendar days after the end of the reference quarter)
RHS: Average revision (first versus current estimate; in percentage points)
90
0.60
80
0.50
70
0.40
60
0.30
50
0.20
40
0.10
30
0.00
20
-0.10
10
-0.20
0
-0.30
00q1
01q1
02Q1
03Q1
04q1
05q1
06q1
8
Summary for GDP revisions
• Euro area revisions are small and estimates are stable
– Size: 0.06ppt (flash), 0.09ppt (annual), 0.12ppt (current)
– Sign: almost negligible (0.01ppt, 0.0 ppt and 0.0ppt)
– Volatility: std of 0.15 for annual update
• Timeliness
– considerable improvements from 70 to 44 days
– Flash estimate did NOT increase revisions, rather opposite
• Infra-annual distribution
– Revisions differ across quarters
– Higher revisions mostly in 1st half of the year
Retail trade
• Flash estimate since January 2004
– Two period: before and after flash
• Flash/first estimate
– versus 1 month later
– versus 2 months later
– versus 1 year later
– versus current estimate
9
Evolution of euro area retail trade growth
3
2.5
Range of estimates
First (flash) estimate
Current estimate
2
1.5
1
0.5
0
-0.5
-1
-1.5
-2
-2.5
-3
Jan.2000
Jan.2001 Jan.2002
Jan.2003
Jan.2004
Jan.2005 Jan.2006
Jan.2007
10
New:Timeliness and average revisions for euro
area retail trade turnover
LHS: Timeliness (number of calendar days after the end of the reference month)
RHS: Average revision (first versus current estimate; in percentage points)
80
2.5
70
1.5
60
50
0.5
40
-0.5
30
20
-1.5
10
0
Jan.2000
-2.5
Jan.2001
Jan.2002
Jan.2003
Jan.2004
Jan.2005
Jan.2006
Jan.2007
11
12
Summary for retail trade revisions
• Periods with rather high (early-2000, turn of 2001; mid2002 until mid-2004) and with smaller revisions
– Size: large before the flash (0.55ppt (annual)) significant
thereafter (0.36ppt (annual))
– Sign: small, but changed direction prior (-) and after (+) flash
– Volatility: decrease since flash from 0.71 to 0.43 (1 year later
estimate)
13
Summary for retail trade revisions (cont’d)
• Timeliness
– considerable improvements from 70 to 35 days
– Initial problems, but flash did NOT increase revisions
• Infra-annual distribution
– No general rule, but pattern differ across months
– Somewhat higher in 1st half of the year
14
Conclusions/caveats
• Revisions in euro area aggregates relatively
small for GDP and significant for retail trade
• Improved timeliness did not lead to higher
revisions
• Infra-annual revisions not equally distributed
• Small sample (from 2000 only)
• No full revision cycle for ‘current’ estimates
• National revisions tend to outweigh each other
• Changes in country coverage not considered
Thank you for your attention
Questions and comments?