Reengineering the Census Bureau's Annual Economic Surveys (2018) / Chapter Skim
Currently Skimming:
5 Sampling and Estimation
5 Sampling and Estimation
Pages 91-118
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 91... ...
Census Bureau's annual economic surveys in a single chapter. Sample design features include • the frame -- that is, the list of units (e.g., enterprises or establish ments)
|
From page 92... ...
The recommendations in this chapter call for the Census Bureau to take three steps: (1) use the same general stratified sampling design, using equal probability sampling within strata, for most or all of the annual economic surveys; (2)
|
From page 93... ...
. The Census Bureau uses stratified sampling for most of the annual economic surveys but with different methods for selecting sample cases within strata.
|
From page 94... ...
A common format will show data users a coherent picture of the whole system and facilitate their understanding of the sampling and estimation methods in the current surveys and, ultimately, in an ABSS. RECOMMENDATION 5-1: The Census Bureau should adopt standardized terminology for sample design features, sampling units, and estimation methods for the annual economic surveys and provide a glossary to users and staff for both external and internal communication.
|
From page 95... ...
In particular, a review would be useful to evaluate whether the goals of the establishment-based surveys can be met by sampling enterprises as the initial units, followed by subsampling establishments within sample enterprises. A two-stage design could be less efficient for ASM and MOPS: if changing their sampling units to enterprises is feasible, however, it would simplify coordination of samples across the annual economic surveys (see Section 5.5)
|
From page 96... ...
TABLE 5-1 Industry Sector Coverage of the Annual Economic Surveys 96 Sector NAICS ASM M3UFO MOPS ARTS AWTS SAS ACES ICTSa ASE SQ-CLASSb COS Agriculture, 11 Partial Partial Partial Forestry, Fishing, and Hunting Mining 21 x x x x x Utilities 22 x x x x x x Construction 23 x x x x Manufacturing 31–33 x x x x x x x Wholesale Trade 42 x x x x x x Retail Trade 44–45 x x x x x x Transportation 48–49 x x x x x x and Warehousing Information 51 x x x x x x Finance and 52 x x x x x x Insurance Real Estate 53 x x x x x x Rental and Leasing Professional, 54 x x x x x x Scientific, and Technical Services Management of 55 x x x x Companies and Enterprises
|
From page 97... ...
appearing on the Business Register each quarter; see Chapter 3. SOURCE: Information provided to the panel by the Census Bureau.
|
From page 98... ...
98 REENGINEERING THE CENSUS BUREAU'S ANNUAL ECONOMIC SURVEYS TABLE 5-2 Strata and Sample Selection Methods Used in the Annual Economic Surveys Sample Survey Sampling Unit Stratification Selection Frame Size Sample Size ASM Establishment NAICS and STPPS 101,000 50,000 (employer) eligible/not eligible to be mailed a questionnairea M3UFO Enterprise NAICS STPPS 72,000 6,000 (in 47 manufacturing industries that historically have had unfilled orders)
|
From page 99... ...
Such a design would facilitate the development of an Annual Business Survey System that uses enterprises as the primary sampling unit. 5.3 SAMPLE SELECTION METHODS Two basic within-strata sample selection methods are used in the 10 annual economic surveys with probability samples.
|
From page 100... ...
could readily be redesigned to use the same method. The panel's preferred method as explained below is equal probability sampling obtained by either simple random or systematic sampling within strata.
|
From page 101... ...
As shown in Table 5-3, the annual economic surveys use a variety of methods for designating certainty units, including employment, inventory, sales, and expert judgment by staff. In a number of surveys, a substantial proportion of the totals of key variables are accounted for by certainty units, especially within small publication cells.
|
From page 102... ...
TABLE 5-3 Methods of Designating Certainty Units and Percentage of Sample and Key Variables Accounted for by 102 Certainty Units in the Annual Economic Surveys Percentage of Sample in Certainty Percentage of Key Variables Survey Certainty Designation Units Accounted for by Certainty Units ASM Establishments with 1,000+ employees Establishments: 50% 87% of sales Top 10 employers in each NAICS6 cell All establishments in each state x NAICS4 cell with number of establishments < 20 Some certainty industries Largest establishments in cost of fuels, purchased electricity, total inventories, total assets, and last-in, first-out inventories M3UFO Arbitrary selections by M3UFO staff of enterprises Enterprises: 31% 71% of sales with large relative measure of size MOPS Same as ASM Same as ASM Not available ARTS Annual sales or end-of-year inventories above industry- Enterprises: 18% 58% of sales specific cutoffs AWTS Annual sales or end-of-year inventories above a Enterprises: 24% Distributors: 52% of sales predetermined cutoff Manufacturer's sales branches and Manufacturer's sales branches offices: 63% and offices: 93% of sales Agents and brokers: 28% Agents and brokers: not available SAS Annual sales or end-of-year inventories above some Enterprises: 25% 65% of sales cutoff ACES 500+ employees Enterprises with employees: 41% Not available Entire sample: 25% ICTSa 500+ employees Enterprises with employees: 38% Not available ASE Based on sales, payroll, or employees Enterprises: 10% Not available
|
From page 103... ...
aSurvey is currently suspended. SOURCE: Information provided to the panel by the Census Bureau.
|
From page 104... ...
This experimentation would have to be carried out for key variables in each survey. RECOMMENDATION 5-4: The Census Bureau should formally investigate a unified approach to determining certainty units in the annual economic surveys by using either the modeling approach described in this panel's report or, if that is not feasible, a standard ized empirical approach.
|
From page 105... ...
Note that the particular units identified as being certainty units may change over time, but the algorithm for determining the certainty cutoff should be stable over time -- a feature that will facilitate operations. If the recommended research leads to the adoption of a unified empirically based approach for designating certainty units in the annual economic surveys, the Census Bureau will need to explain to users the benefits of such an approach and the characteristics of the chosen algorithm.
|
From page 106... ...
Rotation of samples would have the benefit of controlling the burden on some noncertainty units and allow the Census Bureau to inform sample units how long they are obligated to report. Synchronization or coordination is a means of limiting the number of different surveys in which a unit is sampled.
|
From page 107... ...
The surveys included are those that are currently administered annually. SOURCE: Information provided to the panel by the Census Bureau.
|
From page 108... ...
RECOMMENDATION 5-5: The Census Bureau should track the response rates for the annual economic surveys that keep respon dents in the sample for 5–6 years to determine whether the rates decline from year 1 to year 5 or 6. RECOMMENDATION 5-6: The Census Bureau should study the possible benefit of establishing a sample rotation for those of the annual economic surveys that currently draw a sample only once every 5–6 years.
|
From page 109... ...
SAMPLING AND ESTIMATION 109 TABLE 5-5 Types of Estimators, Unit Nonresponse Adjustments, Variance Estimation Methods, and Benchmarking Methods Used in the Annual Economic Surveys Unit Variance Type of Nonresponse Estimation Survey Estimator Adjustment Method Benchmarking ASM Horvitz- None Formula for Benchmarking Thompson Poisson sampling to most recent for capital used even though economic expenditures, sample is now censuses results electricity, Pareto PPS inventory reserve, and purchased services; difference estimator for all other variables M3UFO Horvitz- Not applicable Rosen's variance/ No Thompson covariance for benchmarking estimator, which M3UFO ratio is then ratio components, adjusted to ASM combined with ASM variance as input to a Taylor series approximation MOPS Weighted average Cell based; Stratified Not applicable of management adjustment is jackknife scores for all ratio (count establishments of 2015 ASM in estimation establishments in domain mailout sample that were in tabulations) / (count of 2015 ASM establishments in mailout sample that were in ASM tabulations and satisfied 2015 MOPS response criteria)
|
From page 110... ...
SOURCE: Information provided to the panel by the Census Bureau.
|
From page 111... ...
where Ycen,h is the total value for stratum h from the last economic census, yt,hi is the time period t value for sample unit hi, ycen,hi is the previous census value for the sample unit, and weight whi varies depending on category of establishment. For certainty units and units that do not receive a questionnaire because the Business Register is the source of information on them, the weight is 1; for noncertainty units, the weight is the inverse of the selection probability for a sample unit.
|
From page 112... ...
Another avenue for exploration is whether longitudinal data can be used to a greater extent to reduce variances in the annual economic surveys, as well as providing the basis for analysis of business dynamics. Other surveys, such as BLS's Current Population Survey (CPS)
|
From page 113... ...
Any improvements using longitudinal data would apply only to the noncertainty parts of the samples. RECOMMENDATION 5-7: The Census Bureau should system atically investigate whether auxiliary data can be used more ex tensively in estimation in the annual economic surveys.
|
From page 114... ...
. RECOMMENDATION 5-8: The Census Bureau should study the feasibility of incorporating small-area estimation to produce esti mates for key variables from the annual economic surveys that will better meet users' needs for greater geographic and industry detail.
|
From page 115... ...
Lastly, it will be important to communicate to users the reasons for and extent of differences between preliminary and final estimates. RECOMMENDATION 5-9: The Census Bureau should study the feasibility of producing reliable preliminary estimates of key statistics from the annual economic surveys and, ultimately, from an Annual Business Survey System.
|
From page 116... ...
. RECOMMENDATION 5-10: The Census Bureau should con duct research on what alternative economic data are available and whether there are nonsurvey data sources that can be used to pro duce more timely estimates, more accurate estimates, or both, or otherwise supplement the annual economic surveys and, ultimately, an Annual Business Survey System.
|
From page 117... ...
Department of Commerce Census Bureau. Avail able: https://www.oecd.org/std/30893121.pdf.
|
From page 118... ...
. Sampling methods for repeated business surveys.
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.