The Bicentennial Census New Directions for Methodology in 1990: 30th Anniversary Edition (2015) / Chapter Skim
Currently Skimming:
6 Taking the Census II: The Uses of Sampling and Administrative Records
6 Taking the Census II: The Uses of Sampling and Administrative Records
Pages 215-244
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 215... ...
conducting follow-up of a sample of households that do not return their questionnaires, and (3) implementing coverage improvement programs for hard-to-count areas and population groups on a sample basis.
|
From page 216... ...
Finally, the panel reviewed the uses of sampling in conjunction with administrative records for verification and improvement of the quality of subject items collected in the census. Sampling is also discussed in Chapter 8 in the context of coverage evaluation methods.
|
From page 217... ...
It is, of course, also possible for the sample survey design to introduce nonsampling error. Furthermore, certain components of nonsampling error appear as variances that decrease with increasing sample size.
|
From page 218... ...
The e panel supports further research directed toward evaluating the merits of limited use of sampling as part of the census enumeration process. Taking a Sample Census Currently, decennial census methodology involves collecting the majority of population and housing characteristics from a sample of households, who receive the "long-form" census questionnaire.
|
From page 219... ...
These include projects to develop appropriate sampling error estimates for alternative designs, to develop total error models (including sampling and nonsampling error) , to investigate the theoretical reduction in nonsampling error required to obtain overall accuracy at least equal to that of a complete count, and to develop cost models and estimate their parameters for a sample census.
|
From page 220... ...
Similarly, the field operations of a census, including follow-up and special coverage improvement programs, are geared toward finding every housing unit and person and adding missed units to the address list developed in advance of the census. For a sample census, it is unlikely that the same effort would or could be put into adding units to the sampling frame, and less complete coverage may result.
|
From page 221... ...
These projects are similar to those proposed in connection with conducting the entire census on a sample basis, namely to develop sampling error estimates and total error models for alternative sampling designs. These research endeavors were expected to lead to a pretest of sampling for follow-up and such a pretest was included in the Census Bureau's initial plans for 1986 (Johnson, 1984)
|
From page 222... ...
. Carrying out follow-up operations on a sample basis would also add problems for coverage improvement and coverage evaluation programs that involved matching individual records.
|
From page 223... ...
Census Bureau field staff estimate that as many as 98 percent of households were enumerated by the end of this first stage. The second phase of follow-up included attempts to locate the remaining 1 or 2 percent of nonrespondents and implementation of special coverage improvement programs such as the Vacant/Delete Check and the Nonhousehold Sources P rogram discussed in Chapter 5.
|
From page 224... ...
and, for each stage, determine cost structures and patterns of response, comparing these across different sized geographic areas and areas with differing initial mail response rates. In addition, we suggest that the Census Bureau investigate methods of making the most effective use of field staffs with varying skills and determine if there are new techniques that can be applied to reduce the nonsampling components of total error.
|
From page 225... ...
SAMPLING FOR COVERAGE IMPROVEMENT Along with proposals to follow up a sample of nonrespondents, proposals have been put forward to conduct specific coverage improvement
|
From page 226... ...
If coverage improvement programs are carried out on a sample basis by higherquality staff using careful procedures, it is possible that quality may be improved -- although experience with postenumeration coverage evaluation surveys would not appear to support this hypothesis. On the negative side there are problems of costs and delays in estimation raised by the use of sampling for coverage improvement programs.
|
From page 227... ...
These programs are important for developing a complete list of housing units, which is an essential tool for obtaining complete population coverage. Among the coverage improvement procedures implemented after Census Day, the Vacant/Delete Check stands out as a procedure that: (1)
|
From page 228... ...
t • In 1960, about three-fourths of the population and two-thirds of the housing items were asked on a sample basis. Sample sizes were 25 percent for population items and 25, 20, and 5 percent for hous ing items.
|
From page 229... ...
The Census Bureau ( ohnson, 1984) is also considering conducting a follow-on sample survey J that would collect additional items that are not on either the long form or the short form for about 1-2 million households (1-2 percent)
|
From page 230... ...
For example, the panel is troubled by the proposal to ask questions on noncash income, given that respondents may react negatively and that alternative data sources currently exist for information on this topic, including the new Survey of Income and Program Participation and administrative records. The panel suggests that the Census Bureau articulate explicit criteria for an item's inclusion on the follow-on survey.
|
From page 231... ...
Data from 1980 census returns suggest that a longer short form might not reduce initial response rates ap preciably. Overall, the mail return rate in 1980 for short forms was about 1.5 percentage points higher than the rate for long forms.
|
From page 232... ...
They propose several research projects in this area. Most of their discussion, however, concerns reinterview operations, such as the Vacant/ Delete Check, that are more properly characterized as coverage improvement programs designed to add occupied housing units and persons to the count rather than to change responses to content items.
|
From page 233... ...
For the content improvement programs that appear worthwhile, sampling will often be necessary to make the process manageable in the field and to keep costs within reasonable bounds. Because programs to adjust census reports based on verification or alternative data collection operations have rarely been a part of decennial census methodology, it would be prudent for the Census Bureau to set forth and follow a step-by-step research and testing program.
|
From page 234... ...
However, time constraints precluded examining other items besides housing structure characteristics. The housing items offer the important advantage that concerns over possible invasion of privacy from using administrative records as a data source seen very unlikely to arise in contrast to the use of administrative records to obtain, for example, income data.
|
From page 235... ...
. Nonetheless, investment in research and testing of the use of administrative records for housing structure items offers the potential to improve the accuracy of the data while reducing respondent burden in the census (or, alternatively, permitting other useful items to be put on the questionnaire)
|
From page 236... ...
Figure 6.1 shows several lines plotting costs against sampling rates devel ped from the Census Bureau estimates for a mid-decade census. o These lines indicate that a 50 percent sample survey (the x's on the chart)
|
From page 237... ...
50 A (0.2% rate, $10 million) 0 0 20 40 60 80 100 SAMPLING RATE (percent)
|
From page 238... ...
The calculation of the coefficient of variation includes a factor of 2 for the design effect resultng from sampling entire households rather than conducting a simple random sample i of persons. SOURCE: Calculated from Herriot (1984:Table 1)
|
From page 239... ...
. If the Vacant/Delete Check were also conducted on a 25 percent sample basis, savings for this program might be in the range of: $36 million × (100 – 58)
|
From page 240... ...
First follow-up 8.5 9.7 1 8.5 34.0 4.0 4.5 2 8.0 32.0 2.0 2.3 3 6.0 24.0 1.0 1.1 4 4.0 16.0 Subtotal 15.5 17.6 26.5 106.0 Second follow-up 2.0 2.3 5 10.0 40.0 Total 17.5 19.9 36.5 146.0 NOTE: Number of housing units, number of callbacks, and cost are in millions; 17.5 million housing units is about 20% of the total count of 88 million housing units in 1980. SOURCE: See discussion in Appendix 6.2.
|
From page 241... ...
The impact of this rate and others on the quality of the population estimates for small areas would need to be assessed. We note that the overall rate of contact for the total population of an area using a 25 percent second-stage follow-up sample implemented after two calls in the first stage would be about 95 percent for an area with "average" unit nonresponse of 20 percent, while the rate of contact would be under 90 percent for an area with a 50 percent nonresponse rate.
|
From page 242... ...
Administrative records data for age of structure and other items would then be obtained for the structures in the sample. The sample of basic addresses or structures can be linked to the sample of housing units in the census as follows.
|
From page 243... ...
The v alues obtained from administrative records could simply be appended to the census data records for persons in structures that are in the sample of basic addresses. For persons not in sampled structures, it would be possible to assign values obtained from sampled structures located in the same area.
|
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.