View
27
Download
0
Category
Preview:
DESCRIPTION
Q 2008 European Conference on Quality in Official Statistics Rome 9 July – Early afternoon session: Data Integration I Correction for Coverage Errors in Enterprise Surveys – a Register-based Approach ___________________________________________________________________ - PowerPoint PPT Presentation
Citation preview
Q 2008 European Conference on Quality in Official StatisticsRome 9 July – Early afternoon session: Data Integration I
Correction for Coverage Errors in
Enterprise Surveys
– a Register-based Approach
___________________________________________________________________
Anders Wallgren & Britt Wallgren
Statistics Sweden
• National Accounts (NA) find inconsistencies But difficult to find the causes
• National Accounts (NA) find inconsistencies But difficult to find the causes
• During 2006-2007 a project at Statistics Sweden: Persons from NA, many Business Surveys and the Business Register (BR) worked with this
• National Accounts (NA) find inconsistencies But difficult to find the causes
• During 2006-2007 a project at Statistics Sweden: Persons from NA, many Business Surveys and the Business Register (BR) worked with this
• We found coverage errors and methods to reduce these errors
• National Accounts (NA) find inconsistencies But difficult to find the causes
• During 2006-2007 a project at Statistics Sweden: Persons from NA, many Business Surveys and the Business Register (BR) worked with this
• We found coverage errors and methods to reduce these errors
• The Business Register (BR) will be improved
The old method:
The old method:
• The BR was based on ONE administrative source, BR was a copy of the Tax Authority’s register only
The old method:
• The BR was based on ONE administrative source, BR was a copy of the Tax Authority’s register only
• Sample surveys were based on Business Frames based on the current stock version of the BR
The old method:
• The BR was based on ONE administrative source, BR was a copy of the Tax Authority’s register only
• Sample surveys were based on Business Frames based on the current stock version of the BR
• Register-based surveys were based on the Business population in each administrative source
The old method:
Frames used for yearly NA year 2004:
Yearly framefor SBS 2004
Quartly frames
Deliveries of administrative data for 2004
2004 2005 2006
The new idea:
New kind of Business Register is created January t+2:
Yearly framefor SBS 2004
All enterprises active during 2004Quartly frames
NA year 2004Deliveries of administrative data for 2004 data deliviery
Calendar year register 2004based on ALL sources
2004 2005 2006
Coverage errors with the old method:
Number of Legal units (LU)
November frame 2004 Calendar year register (CYR) 2004
47 662 "Has never been active" acc. to Nov. frame88 227 "Not active" according to Nov. frame
171 688 Missing completely in Business Register307 577 Total undercoverage
15 Nov. 2004 15 J an. 2006
779 277
307 577
Not in CYR
2005
93 114"Active" according to Nov. frameOvercoverage
"Active" according to November frame In CYR
In CYR
779 277
Coverage errors with the old method:
SEK, Millions
Errors due to undercoverage in the November frame 2004Wage sum Turnover
SEK Millions SEK Millions
"Active" according to the November frame 1 003 186 5 582 374
"Has never been active" according to November frame 177 1 734
"Not active" according to November frame 5 671 112 363
Missing in November frame 639 5 061
Totalt in Calendar Year Population 1 009 673 5 701 532
Errors due to undercoverage in the November frame 6 487 119 158
Errors due to undercoverage in the November frame, % 0.65 2.13
Coverage errors with the old method:
SEK, Millions
Undercoverage in November frame 2004 by industry
Wage sums, LSUM-register Turnover, VAT-register Non financial sector Total Undercoverage Total Undercoverage
Industry SEK
millions SEK
millions Percent
SEK millions
SEK millions
Percent
01 4 663 341 7,3 64 720 4 425 6,8 10+11+12 178 14 8,0 1 399 24 1,7
212 2 988 2 0,1 26 094 4 441 17,0
23 1 040 1 0,1 11 761 1 455 12,4
702 except 01 4 661 338 7,2 95 009 3 536 3,7
91 1 776 14 0,8 5 327 801 15,0 … … … … … … …
Totalt: 650 102 5 629 0,9 5 453 021 116 324 2,1
Coverage errors with the old method:Overercoverage in November frame and Business Structure Survey 2004 by industry
Value of production total Overcovaerage Overcoverage
Industry SEK
millions No of legal
units SEK
millions No of legal
units % SEK
% legal units
01 62 940 130 283 1 110 6 787 1.8 5.2
02 50 430 42 022 849 2 726 1.7 6.5
05 1 204 1 514 43 106 3.5 7.0
273 4 990 76 74 2 1.5 2.6
362+363 815 906 17 87 2.1 9.6
502 17 356 10 857 259 787 1.5 7.2
71 21 968 5 458 324 349 1.5 6.4
741 77 787 55 123 1 705 5 570 2.2 10.1
742+743 64 678 29 892 972 2 890 1.5 9.7
745-748 77 137 31 130 1 294 4 098 1.7 13.2
851 43 417 19 358 1 687 1 543 3.9 8.0
852 1 796 921 85 54 4.7 5.9
91 6 445 1 364 187 82 2.9 6.0
93 13 483 27 617 351 2 260 2.6 8.2 … … … … … … … Summa 3 554 910 777 793 21 604 59 773 0.6 7.7
Coordination between different surveys:
Legal units (LU)Number of legal units in the Calendar Year Register for 2004 by sector and industry
Industry Sector 01-64 65 66 67 70-99 Total 1 184 889 1 687 35 1 915 161 401 349 927 2 11 761 1 049 1 788 18 3 627 3 28 10 1 1 376 416 4 12 5 0 0 445 462 6 407 408 47 2 371 191 159 598 987 7 1 352 709 75 12 26 450 28 598 Total 593 700 3 219 1 162 4 087 379 849 982 017
Coordination between different surveys:
Legal units (LU)
BSS / Financial survey: Overlap ~ 100 LU
BSS / Financial survey: Gap ~ 6 000 LU
Number of legal units in the Calendar Year Register for 2004 by sector and industry
Industry Sector 01-64 65 66 67 70-99 Total 1 184 889 1 687 35 1 915 161 401 349 927 2 11 761 1 049 1 788 18 3 627 3 28 10 1 1 376 416 4 12 5 0 0 445 462 6 407 408 47 2 371 191 159 598 987 7 1 352 709 75 12 26 450 28 598 Total 593 700 3 219 1 162 4 087 379 849 982 017
Inconsistent frame populations:
Business Structure Survey / Energy survey
Business Structure Survey (BSS) and Energy Statistics (EN).
Different identity numbers are included in the two frame populations.
Some identity numbers are included in both, others in only one of the surveys.
Income 1998 as per : Income 1999 as per :
BSS EN BSS EN
Only BSS 32% 0% Only BSS 16% 0%
Both 68% 76% Both 84% 96%
Only EN 0% 24% Only EN 0% 4%
Total 100% 100% Total 100% 100%
Business Structure Survey (BSS) and Energy Statistics (EN).
Different identity numbers are included in the two frame populations.
Some identity numbers are included in both, others in only one of the surveys.
Income 1998 as per : Income 1999 as per :
BSS EN BSS EN
Only BSS 32% 0% Only BSS 16% 0%
Both 68% 76% Both 84% 96%
Only EN 0% 24% Only EN 0% 4%
Total 100% 100% Total 100% 100%
Conclusion:
• When ALL administrative sources for year t have been delivered:
Conclusion:
• When ALL administrative sources for year t have been delivered:
Create the Calendar Year version of the BR with all enterprises active during some part of year t
Conclusion:
• When ALL administrative sources for year t have been delivered:
Create the Calendar Year version of the BR with all enterprises active during some part of year t Sector and Industry are defined by this register
Conclusion:
• When ALL administrative sources for year t have been delivered:
Create the Calendar Year version of the BR with all enterprises active during some part of year t Sector and Industry are defined by this register
• The population defined by this register is used by ALL sample surveys and register-based surveys that deliver data to the yearly National Accounts
• Sample surveys are based on earlier frames The preliminary estimates are calibrated to be consistent with the Calendar Year Register
• Sample surveys are based on earlier frames The preliminary estimates are calibrated to be consistent with the Calendar Year Register
CYR - Calendar Year Register year tRegister population: All enterprises active in at least one administrative source for year t
Economic variables, SEK millionsAdministrative sources Sample surveys
BIN Sector Industry Turnover Wage sum Survey 1 Survey 2
1 1 12345 250 1002 1 12346 150 60 433 1 12347 100 40 144 1 12348 50 20 185 1 12349 100 30 226 1 12340 120 407 1 12341 80 20•••N
Consequences for the NA:
Consequences for the NA:
• Small undercoverage and overcoverage errors
• Small overlap / gaps between surveys
Consequences for the NA:
• Small undercoverage and overcoverage errors
• Small overlap / gaps between surveys
Consistent estimates!
Recommended