Supporting Statement B (1220-0050)

Supporting Statement

B. Collection of Information Employing Statistical Methods

1. Consumer Units
   
   

There are approximately 116.4 million consumer units (CUs) in the potential CE survey universe.¹ A CU is the unit from which we desire expenditure reports. It consists of all household members of a particular housing unit or other type of living quarters who are related by blood, marriage, adoption, or some other legal arrangement. The CU determination for unrelated persons is based on financial independence. The unrelated persons are considered separate CU(s) if they are responsible for paying their own expenses for two out of three of the following expense categories: shelter, food, and all other expenses. Approximately 97 percent of all occupied living quarters are a single CU.

For an overview of the CE sample design and the CU selection process, please refer to the 2007 CE Anthology article, "How Was My Household Selected for the Consumer Expenditure Survey? The Design and Selection of the Survey's Sample" by Sylvia Johnson-Herring and Susan King. This article is available upon request.

The following table shows estimated numbers of CUs for all 91 strata from which PSUs were selected.² (See the section below entitled “PSUs” for more information.)

Stratum Code	Estimated Number of CUs in Stratum
A102	2,559,213
A103	2,935,498
A104	1,403,563
A109	3,311,790
A110	3,192,066
A111	2,754,939
A207	3,793,086
A208	2,256,483
A209	1,325,561
A210	1,218,236
A211	1,296,962
A312	2,079,224
A313	1,055,780
A316	2,181,835
A318	1,950,037
A319	1,737,397
A320	1,603,061
A321	1,313,016
A419	5,113,753
A420	1,657,499
A422	2,911,099
A423	1,470,056
A424	1,163,649
A425	1,212,368
A426	1,068,366
A427	1,044,510
A429	1,344,800
A433	1,309,891
X102	1,482,049
X104	1,275,807
X108	1,348,185
X210	1,075,910
X212	1,082,608
X218	1,128,354
X220	1,128,308
X222	1,059,808
X224	1,082,708
X226	1,137,025
X228	946,721
X232	1,392,266
X336	1,287,497
X338	1,221,154
X340	1,398,206
X342	1,550,352
X344	1,281,584
X346	1,276,391
X350	1,114,492
X352	1,247,393
X354	1,193,368
X356	1,260,701
X358	1,238,172
X362	1,346,115
X364	1,367,570
X366	1,326,273
X368	1,219,167
X472	1,090,016
X474	1,081,970
X482	1,145,328
X484	1,164,280
Y102	562,716
Y104	654,250
Y206	825,058
Y208	981,226
Y210	817,980
Y212	980,953
Y314	812,774
Y316	938,987
Y318	778,777
Y320	819,967
Y322	969,428
Y324	776,087
Y426	528,041
Y428	482,697
Y430	595,002
Y432	575,530
Z102	369,878
Z104	575,351
Z206	849,611
Z208	624,732
Z210	768,195
Z212	912,352
Z314	773,198
Z316	733,707
Z318	777,472
Z320	1,020,768
Z322	943,804
Z324	561,559
Z426	271,401
Z428	235,813
Z430	325,837
Z432	343,333
Total	116,400,000

Response Rates

The following table shows expected response rates for the Quarterly Interview and Diary Surveys based on 2005 response rates.

Category	Quarterly (quarter)	Diary (annual)
Total Sample Size	14,600	12,100
Total Type B and C Noninterviews (vacant, demolished, etc.): Number Percent of Total Sample	2,701 18.5	2,238 18.5
Total Eligible Units	11,899	9,862
Total Type A Noninterviews Number Percent of Total Eligible	3,074 25.8	2,812 28.5
Total Completed Interviews Number Percent of Total Eligible	8,825 74.2	7,050 71.5

Note: For more information on the calculation of response rates, please see “Updating the Monthly Nonresponse Reports,” a document maintained in the Demographic Surveys Division of the Bureau of the Census.

PSUs

The CE Survey uses a multistage stratified sample. The set of sample PSUs used for both the Quarterly Interview and Diary Surveys consists of 91 PSUs, 75 of which the BLS defined and selected for the CPI. The BLS derived these PSUs from a stratified design with the objective of selecting one PSU from each stratum. Stratification in the X, Y, and Z PSUs used a 5-variable geographic model whose independent variables are normalized longitude, the square of normalized longitude, normalized latitude, the square of normalized latitude, and percent urban consumer units.

The 91 CE PSUs fall within four categories:

PSU Letter	Definitions	Number of PSUs
A	Self-Representing Metropolitan CBSAs	21
X	Non-Self-Representing Metropolitan CBSAs	38
Y	Micropolitan CBSAs	16
Z	Non-CBSA PSUs	16

2. Sampling Within PSUs

To select living quarters within each PSU, the Census Bureau uses a four-frame approach accessing a combination of Census 2000 address lists, a specially constructed and continuously updated frame of new construction permits, and area segments for the sample selection. (The four frames are often called Unit, Area, GQ, and Permit.) The Area frame consists of units in blocks that are either in non-permit-issuing areas, or in areas for which the Census Bureau does not have reliable address list. Group quarters in Area frame blocks are automatically sampled along with the other units during Area frame listing operations. New construction in Area frame blocks is listed along with old construction; however if a unit is selected for sample from an Area frame block in a permit-issuing area, the interviewer asks what year the unit was built before beginning the interview. If the unit was built after 1999, then the unit is dropped from the Area frame sample because it has a chance of selection in the Permit frame. The Group Quarters (GQ) frame consists of GQ units in non-Area frame blocks which existed at the time of the 2000 Census. The Unit frame consists of non-GQ units in non-Area frame blocks which existed at the time of the 2000 Census. The Permit frame consists of permit new construction units in permit-issuing areas.

For more information on sampling within PSUs for the CE surveys, please refer to "Determining Within-PSU Sample Sizes for the Consumer Expenditure Survey" by David Swanson, Sharon Krieger, and Sylvia Johnson-Herring in the Proceedings of the Section on Government Statistics, American Statistical Association, 2003. This article is available upon request.

In general, since the sample sizes are small, we are using segments with clusters of size one to achieve maximum reliability of the estimates.

The Census Bureau uses an un-clustered systematic sample design to select the Unit frame sample from the 2000 census 100 percent detail file. The sample drawn from the detail file is augmented by sample from the new construction (Permit) frame. This frame consists of building permits which were authorized for housing units since the Census. A sample of newly constructed housing units represented by the building permits is drawn in each CE PSU.

The sample design for the unit frame stratifies and sorts HUs within an area with living quarters further subdivided into two categories: occupied HUs at the time of the census, and vacant HUs. The design provides for further stratification of occupied HUs by two 2000 census variables: a tenure-value-rent combination and household size. These variables were chosen following extensive research on within-PSU sampling prior to the sample redesign based on the 1990 Census. Because value and rent were only collected from a sample of housing units (the Long Form) in Census 2000, these two variables were imputed for those units not in the (Long Form) sample. Another change from the previous sample redesign is that rent and value categories were based on percentiles within counties, rather than pre-determined cutoffs. A single digit code was determined by tenure-value-rent. Then a second single digit code was determined based on the household size and the value of the first code. The concatenation of these two codes is the CE Unit frame stratum code.

Tables 1 and 2 provide the group definition (stratum code) of each variable and the associated group numbers for occupied HUs.

Table 1: Tenure-Value-Rent Code [Digit 1 of Unit Frame CE Stratification Code]
Tenure	Value/Rent Percentile Range	Group #
Owner	I (0 % - 24 %)	3
	II (25 % - 49 %)	4
	III (50 % - 74 %)	7
	IV (75 % - 99 %)	8
Renter	I (0 % - 24 %)	1
	II (25 % - 49 %)	2
	III (50 % - 74 %)	5
	IV (75 % - 99 %)	6

Table 2: Household Size Code [Digit 2 of Unit Frame CE Stratification Code]
Group # for Tenure-Value-Rent Code	HU Size	Group # for HU Size
1, 3, 5, or 7	1	1
	2	2
	3	3
	4+	4
2, 4, 6, or 8	1	4
	2	3
	3	2
	4+	1

The tenure-value-rent code and HU size code form the first and second digits, respectively, of the stratum code.

Table 3 below provides the strata definition and the corresponding strata codes for the vacant HUs:

Table 3: Unit Frame CE Stratification Code for Vacant HUs
Tenure	Value/Rent Percentile Range	Code for:
		HU: Vacant for Sale/Rent	HU: Residual Vacant
Owner	I (0 % - 24 %)	30	99
	II (25 % - 49 %)	45
	III (50 % - 74 %)	70
	IV (75 % - 99 %)	85
Renter	I (0 % - 24 %)	10
	II (25 % - 49 %)	25
	III (50 % - 74 %)	50
	IV (75 % - 99 %)	65

The Census Bureau forms Basic PSU Components (BPCs) from the intersections of the various demographic surveys’ stratification PSUs. (In the 2000 Sample Redesign, a BPC always consists of one or more counties.) The Census Bureau sorts the CE Unit frame in a given BPC by Urban/Rural classification, FIPS County code, CE Unit frame stratification code, Census Tract code, Census Tabulation Block code, Basic Street Address, and the Unit Sort Order³ code. A systematic sample is then drawn from the sorted file.

The Area, GQ, and Permit frames are constructed from “measures”, where each measure is expected to contain four units. Measures are sorted geographically in the Area and GQ frames, and by Permit Date and Building Permit Office code in the Permit frame. Systematic samples of measures are selected independently in each frame. We expect the distribution of designated sample units among the four frames to be:

Unit: 80%

Area: 10%

Permit: 9%

GQ: 1%

Note, however, that the proportion of sample in the Area and Permit frames will increase as new construction is added to the sample.

Estimation

The estimation procedure for both the Diary and Quarterly Interview Surveys follow

well-known statistical principles in that the final weight for each sample CU is the product of the inverse of the probability of selection; a weight adjustment to account for noninterviews; and a regression-based adjustment that incorporates post-stratifications to adjust for population undercount at the same time controlling for extreme weights. For additional information on the sample design and estimation methodology used in the CE surveys, please refer to "Chapter 16, Consumer Expenditures and Income" in the BLS Handbook of Methods. This chapter is available upon request.

The BLS requires an average coefficient of variation of 10 percent or less covering 2 years of data for Expenditure Classification expenditures for each of the 39 market basket areas.

For additional information, please refer to Kenneth V. Dalton’s memo to Chester E. Bowie, “Specifications for the Selection of CE/CPI Samples in PSUs Based on the 2000 Census:, June 28, 2002 and Alan R. Tupek’s memo to Kenneth V. Dalton, “Calculations of Within-PSU Sampling Intervals for the Census 2000-Based Redesign of the Consumer Expenditure Surveys and the CPI Permit New Construction Housing Sample”, November 11, 2002. These documents are available upon request.

3. In the CE Survey, keeping the noninterview rate at a low level requires special efforts, particularly from the Census Bureau Field staff. For each refusal case, the regional office sends a special letter to the address and assigns the case for follow-up by the program supervisor, supervisory field representative, or senior interviewer, taking into account time and cost considerations.

To adjust for those noninterviews that the field staff cannot convert to interviews, the sample design provides for a noninterview adjustment in the estimation procedure. The computer processing employs special techniques in the Quarterly Interview Survey to reference data provided in the previous interview, keeping recall problems and interview time to a minimum.

4. At the time of this submission, there are no plans to test any procedures or methods in FY2007.

5. The Census Bureau will collect the data. Within the Census Bureau, you may consult the following individuals and their area of expertise for further information.

Sample Design: Karen King (301) 763-1974

Data Collection: Howard McGowan (301) 763-5342

1¹ The number of CUs comes from dividing the Census Bureau’s 2005 estimate of the number of people in the civilian non-institutional population (291 million) by the average number of people per CU (2.5).

2² The number of CUs per stratum comes from allocating the nationwide total of 116.4 million CUs by each stratum’s proportion of the nationwide population in the 2000 Census.

3 The Unit Sort Order code was devised by the Census Bureau in an attempt preserve geographic proximity between units in the sort.

2