Transcript Survey Design Considerations for Supporting Industry-by

27th Voorburg Group Meeting
Warsaw, Poland
Operational Efficiency Strategies Session
(Survey Design Considerations for Supporting Industry-byProduct Estimates for the Service Annual Survey)
presented by
Mark E. Wallace
Chief, Service Sector Statistics Division
U.S. Census Bureau
[email protected]
October 1–5, 2012
Agenda
•
•
•
•
•
Introduction, Background, and Challenge
Meeting the Challenge - 3 Different Phases
Outcome
Conclusion and Next Steps
Questions
2
Introduction, Background, and
Challenge
– Requests for industry-by-product detail for service industries
since the 1980s, primarily from the Bureau of Economic
Analysis (BEA)
– The 2007 Economic Census included the first-time collection of
a rich variety of service industry-by-product detail
– Expansion of industry coverage in the SAS (began for 2009 and
expanded industry-by-product coverage began for 2011)
– Pre-expansion and Post-expansion service industries in the
Service Annual Survey
3
Pre-Expansion and Post-Expansion Service Industries in the
Service Annual Survey
4
Pre-Expansion and Post-Expansion Service Industries in the
Service Annual Survey (continued)
Total of Pre-expansion NAICS (30% of GDP) and Post-expansion NAICS (25% of GDP) = 55% of GDP
5
Introduction, Background, and Challenge (continued)
– It was necessary for the Census Bureau and BEA to
determine priorities for specific industry-by-product
combinations due to the availability of limited resources
– Sample selection for prior SAS samples (e.g., 2000, 2006)
did not account for industry-by-product combinations,
resulting in:
• High CVs
• High imputation rates
– 2007 Economic Census included 2,000 new products
across more than 350 service industries and would serve as
important input into developing potentially collectable and
publishable industry-by-product combinations for the 2011
SAS onward
– Funding for current service expansion received in 2009
made it possible to carry out the expansion
– New sample for SAS selected for 2011 data year, with work
beginning in 2009
6
Introduction, Background, & Challenge (continued)
– Design a sample that will meet target CVs
– For the first time use product CVs in the sample
design
– Design a sample for products & industries never
collected in the SAS
– Limit sample size to approximately 70,000 units to
accommodate resource and funding constraints
7
Meeting the Challenge
3 Distinct Phases
Phase 1: Working with BEA to get initial list of
industry-by-product priorities prior to
feasibility research
Phase 2: Researching the feasibility of BEA’s
industry-by-product priorities
Phase 3: Working with BEA to get a final list
of industry-by-product priorities
8
Meeting the Challenge (continued)
Phase 1-Working with BEA to Get Initial List of Industryby-Product Priorities Prior to Feasibility Research
•
Obtaining a list of industry-by-product combinations BEA was requesting for
the pre-expansion industries and for the post-expansion industries
•
Initially, obtained additional industry-by-product combinations without
identifying any existing industry-by-product combinations that could be
eliminated from the SAS forms (due to non-reportability, high coefficients of
variation, high imputation rates, or lack of economic significance-not enough
companies could report a specific product)
•
Begin focus on pre-expansion industry-by-product combinations, and
determine whether they would require additional sampling units (for those
that had CVs higher than 20%)
•
See Attachment B of paper for the industry-by-product combinations
identified during Phase 1 for the pre-expansion industries, along with an
indication of whether they would require additional sampling units
9
Excerpt of Attachment B from Paper: Industry-by-Product Combinations
Identified During Phase 1 for NAICS 51721, Wireless Telecommunications
Carriers (except Satellite)
Business+Government Revenue
Name
Products
that require
additional
sampling
units
51711
NAICS Level
(2007 basis)
1761+1762
Product/
Special
Inquiry
Keycodes will be changed to
Notes
breakout class of customer for cable
revenue
C
Collect
(C)
or TBD
51721
6019
Internet Access Services
51721
6021
Install. Serv. Telecom Networks
51721
6022
Resale of Telecom Equipment
51721
6023
Rental of Telecom Equipment
51721
6024
Repair/Maintenance Telecom Equip.
C
51721
6027
Mobile Telephony
C
51721
6301
Messaging Services
X
C
51721
6311
Mobile Long Distance
X
C
51721
6312
Mobile All Distance
51721
1761+1762
51721
1763
C
X
TBD
C
X
C
C
Business+Government Revenue
X
C
Household Revenue
X
C
10
Meeting the Challenge
Phase 2 – Researching the Feasibility of BEA’s
Industry-by-Product Priorities
• Researching reportability of products for pre-expansion
industries (e.g., contacting companies for record keeping and
reporting practices
• Using industry-by-product data from the Economic Census
• Using prior (2005-2008) SAS estimates for pre-expansion
industries
• See Attachment C of paper for the industry-by-product
combinations previously identified by BEA in Phase 1 (see
Attachment B) that were subsequently designated to be dropped
• Examining products for post-expansion industries
• See Attachment D of paper for industry-by-product combinations
that could be added for post-expansion industries without
requiring additional sampling units
11
Meeting the Challenge
Phase 3 – Working with BEA to Get a Final List of
Industry-by-Product Priorities
• Discuss results of Phase 1 and Phase 2 with BEA
• Present “Cost” of pre-expansion industry-by-product
combinations where sample size would need to
increase
• Provide a mechanism for BEA to prioritize industryby-product combinations to be included in the
Service Annual Survey
• See Attachment E of paper for the spreadsheet tool
provided to BEA by the Census Bureau for
prioritizing industry-by-product combinations to be
included in the Service Annual Survey
12
Phase 3 - Mechanism for BEA Making Selections
TABLE E
Budget = 4000
Cumlative Total =
Products for BEA's Selection
NAICS
51111
51113
51114
51121
5121X
51219
PRODUCT(S)
DESCRIPTION
Status = OKAY
BEA CHOICE - 1
or BLANK
SELECTED COST
COST
6066 Online Revenue
6103 General Reference Books
6135 Rental or sale of mailing lists
6005 Application Service Provisioning
6195 Outright Sale of AV Works
6222 Motion Picture Film Lab Services
2850
50
80
10
80
900
55
1
120
175
55
190
80
25
1050
930
460
65
2450
165
55
355
770
1
1
1
1
1
1
1
1
1
50
0
10
0
900
55
Admin. of Copyrights of Others; Licensing of Rights-Musical
Compositions; Licensing of rights-musical recordings
51223 6231, 6008, 6009
515112
6254 Public and Non-Commercial Prog.
51512
6254 Public and Non-Commercial Prog.
51711
6023 Rental of Telecom Equipment
51721
6021 Install. Serv. Telecom Networks
51919
6391 Info. Search/Retrieval Services
5412
3070 Computerized Accounting Sys Serv.
54131
3081 Single Family Residential Projects
54133
3111 Residential Engineering Projects
5417E
3208 Licensing of Rights to Use Int. Prop.
5417T
3209 Original Works of Int. Property
54181
3225 Direct Marketing
56131
3264 Long-Term Staffing
5615X
3285 Commissions: Cruises
5629X
3363 Sale of Nonhaz Recyclable Material
Budget = 4000
DUE DATE FOR BEA TO SEND TO CENSUS - AUGUST 13, 2010
1
1
1
Cumlative Total =
Status = OKAY
120
175
55
190
80
25
0
0
0
65
0
0
0
355
770
2850
13
Outcome
• BEA’s choices resulted in being 1150 sampling
units under budget
• Extra available sampling units allowed us to
expand the sample size for a few health
industries, resulting in the Center for Medicare
and Medicaid Studies receiving estimates of
source of funding for additional industries
14
Conclusion and Next Steps
•
•
•
The 2012 Business Sample Revision achieved its goal of considering
industry-by-product combinations in the design of the SAS sample
A carefully planned and implemented three-phase approach produced
an outcome that maximized the usefulness of SAS industry-by-product
data for BEA, while maintaining sustainable workload and resource
levels
Next steps will be to:
– Evaluate the results of the 2011 SAS
– For expansion industries, determine what possible additional industry level
estimates to provide to BEA
– Make necessary adjustments to future SAS forms based on:
•
•
•
•
Ongoing BEA needs
Reporting issues
Consistency with the Economic Census products
Sample size & burden considerations
– Research how to incorporate the 2012 Economic Census results into future
sample designs
15
Questions
16