supporting statement part B.wpd

B. Collection of Information Employing Statistical Methods

1. Universe and Respondent Selection

There are approximately 270 million individuals in the potential SIPP respondent universe (civilian noninstitutionalized population based on the 2000 decennial census). The SIPP uses a multistage-stratified sample of the U.S. civilian noninstitutionalized population. The first stage involves the definition and division of the U.S. into groups of counties called the Primary Sampling Units (PSUs), which are assembled into homogeneous groups called strata. The second stage involves selection of units within the PSU.

The 2004 SIPP sample is the first sample for the SIPP to be selected from the 2000 redesign of the sample. There are 351 PSUs in the 2000 redesigned SIPP and in sample for the 2004 Panel. The selected PSUs cover both urban and rural areas of the U.S. Many of the PSUs in the 2000 SIPP sample design were also in the 1990 SIPP sample design.

Within each PSU, the bulk of the living quarters (LQs) are divided into clusters ranging in size from one to four. Systematic samples of these clusters are selected from lists of addresses prepared for the 2000 decennial census. Other sampling techniques are used to represent new construction, group quarters, areas where incomplete addresses are common, and LQs that were not found in the 2000 decennial census. We oversampled low income households from the lists of addresses prepared for the census and from the lists of incomplete addresses. Wave 1 of the 2004 Panel sample consisted of approximately 62,000 designated LQs, of which approximately 49,600 will be occupied at the time of interview and about 46,500 will be interviewed. Each household contains, on average, 2.1 eligible respondents; therefore, there will be approximately 98,000 Wave 1 survey respondents.

The 2004 SIPP sample of 46,500 households is selected from the SIPP's PSUs in the U.S. Definition of the PSUs, address lists, and area segments are all based on the 2000 decennial census.

Due to budget constraints, the 2004 Panel sample has been reduced by approximately 53% to 21,292 households eligible for interview beginning in Wave 9, October 2006.

2. Procedures for Collecting Information

The raw sample data will be weighted to independent population controls by age, race, household type, sex, and Hispanic¹ origin categories. After the weighting takes place, estimates, such as the number of people below the poverty level, number of social security recipients, and so forth, can be derived directly from the resulting database. The estimation process will also include derivation of sampling errors which is required to test the statistical validity of the survey estimates.

The precision of an estimate is shown by the coefficient of variation (CV), which is the standard error expressed as a percentage of the estimate. With the proposed sample design, statistics, such as the percentage of people in poverty (12 percent), are expected to have a CV of about 2 percent. We view the proposed sample design as the minimum required to provide reliable estimates for a variety of characteristics at the national level.

In sample households, all people 15 years old and over will be interviewed using regular proxy-respondent rules. In accordance with the longitudinal rules of the survey, every effort will be made to interview in Waves 2 through 12 the same people interviewed in Wave 1. Sample individuals (all household members present at the first interview) who move within or reasonably close to the boundaries of the PSUs from which the 2004 SIPP sample was selected will be followed and interviewed. Individuals 15 years old or over who enter the household after Wave 1 will be interviewed. The interviewing cycle for each wave (interview period) of the 2004 SIPP

Panel is as follows:

Wave 1: February 1 through May 31, 2004.

Wave 2: June 1 through September 30, 2004.

Wave 3: October 1, 2004 through January 31, 2005.

Wave 4: February 1 through May 31, 2005.

Wave 5: June 1 through September 30, 2005.

Wave 6: October 1, 2005 through January 31, 2006.

Wave 7: February 1 through May 31, 2006.

Wave 8: June 1 through September 30, 2006.

Wave 9: October 1, 2006 through January 31, 2007.

Wave 10: February 1 through May 31, 2007.

Wave 11: June 1 through September 30, 2007.

Wave 12: October 1, 2007 through January 31, 2008.

The sample is systematically divided into quarters across the 4 months (rotations) of each wave.

A reinterview questionnaire is administered to approximately 3,100 households in Waves 1-12. The reinterview checks for survey falsifications. Households in reinterview are not eligible again for reinterview until the next calendar year. All Type A noninterviews are ineligible for reinterview. Whenever possible, reinterviews will be performed by telephone.

3. Methods to Maximize Response

In all SIPP Panels, we make special efforts to minimize noninterviews. In each Wave, every household in the active sample receives an advance letter that explains the purpose of the survey and why their cooperation is important. In Wave 1, each household is given a portfolio that contains reports created from the SIPP data and brochures informing respondents about the Census Bureau and our commitment to confidentiality. Each refusal case is sent a special refusal letter, and the case is assigned for a refusal follow-up by the program supervisor, senior FR, or an experienced FR. During the 1996 and 2001 SIPP Panels, the Census Bureau experimented with giving monetary incentives to households in order to maximize response. Based on results from these experiments, we plan to use incentives in the 2004 Panel. During Wave 1, the respondents were sent the advance letter in a Priority mail envelope. Starting in Wave 1 through Wave 9 of the 2004 Panel, 100 percent of the survey households were eligible for a debit card worth $40. An incentive was offered to an eligible case if there were signs that the case was clearly planning to leave the survey. At the time the incentive was offered, it was made clear that payment was conditioned on a complete response to the questionnaire. In response to an OMB request to continue to study the SIPP attrition issue, we sent a special letter to households that had refused to participate for two consecutive waves (Waves 4 and 5). These households were informed that in return for their completed interview a debit card would be mailed to the household.

To adjust for those noninterviews that cannot be converted, a noninterview adjustment will be performed in the estimation procedure. Within the instrument, techniques are used that reference data provided on the prior interview to lessen recall problems and keep interview time to a minimum.

4. Tests of Procedures

The 1995 Dress Rehearsal was an important procedural test in preparation for future SIPP/CAPI Panel surveys. The dress rehearsal examined both the revised

questionnaire design and automated collection methodology. Also, we tested the reinterview questionnaire to ensure that it integrated well with the rest of the survey data collection system. The dress rehearsal represented the final stage of an extensive series of cognitive review and instrument tests in the CAPI environment. The use of a complete CAPI environment is based on numerous tests using automated technology at the Census Bureau.

5. Contacts for Statistical Aspects and Data Collection

The Census Bureau will collect and process these data. Within the Census Bureau, the following individuals should be consulted for further information on their areas of expertise:

Sample Design

Tracy Mattingly Chief, SIPP Branch

Demographic Statistical Methods Division

301-763-6445

Data Content

Robert Kominski Assistant Division Chief, Housing and Household Economic Statistics Division

301-763-2120

David Johnson Chief, Housing and Household Economic Statistics Division

301-763-6443

Data Collection and Tabulation

Patrick Benton Chief, Income Surveys Branch

Demographic Surveys Division

301-763-4618

Attachments

A. SIPP-241005(L)–Director's Letter

B. SIPP-24003–Reminder Card

1Hispanics may be of any race.