SupportingStatement_PartB

SUPPORTING STATEMENT

U.S. Department of Commerce

U.S. Census Bureau

Annual Social and Economic Supplement

to the Current Population Survey

OMB Control Number 0607-0354

Part B - Collections of Information Employing Statistical Methods

Question 1. Universe and Respondent Selection

The Basic CPS universe is 114 million households. From the universe of 114 million, we select a sample of approximately 60,500 households. With the SCHIP general expansion, this increased to 72,000 households each month. Of these, approximately 60,000 households will be eligible for interview; and we will actually interview approximately 55,000 households.

The ASEC sample expands upon the Basic CPS sample by adding select households. Approximately 6,000 Hispanic households interviewed in the previous November CPS will be added. We expect to interview approximately 4,500 of these households. Additionally, we will add approximately 8,800 minority and White (with children) households that were interviewed in November. We expect to interview about 6,500 of these households. Finally, we will conduct the ASEC to selected minority and White (with children) households during February and April. These households will be “borrowed” from the February outgoing and the April incoming rotation groups. We expect to select approximately 12,800 such households, with about 12,000 actually being interviewed. This brings us to a total of approximately 78,000 households planned for interview in fiscal year 2014.

We use a household respondent to answer the supplement items for all household members 15 years of age or older. The response rate for the ASEC averages 90 percent.

Question 2. Procedures for Collecting Information

This is a supplemental survey associated with the CPS. Attachment B gives an overview of the CPS sample, design, weighting methodology, and response rates. The statistical properties of these supplemental items will fall within those associated with the CPS itself.

Question 3. Methods to Maximize Response

Response rates and data accuracy for the CPS are maintained at high levels through computer edits, interviewer instruction and training, and close monitoring of the data. Refer to "Overview of CPS Design and Methodology," item 5, for a discussion of CPS nonresponse (Attachment B).

Some amount of allocation for item nonresponse is done within the ASEC supplement for eligible people. Item nonresponse rates range from less than

2 percent for the work experience items to more than 15 percent for some of the income items, such as interest and dividends. For households, the basic CPS household-level nonresponse rate was 10.72 percent. The household-level nonresponse rate for the ASEC was an additional 10.8 percent. These two non-response rates lead to a combined supplement nonresponse rate of 20.4 percent.

Question 4. Tests of Procedures or Methods

Recent studies have shown the ASEC yielding different estimates of varying magnitudes across income types, health coverage, and particular subpopulations, when compared to other data sources. These differences typically lead to under-reporting of statistics. For example, enrollment in public benefits programs (such as food stamps) is often under-reported, even after imputations for missing data are performed. Similarly, under-reporting of health coverage has also been a persistent problem.

In an on-going effort to curb under-reporting, Westat, Incorporated (referred to hereafter as Westat) and the Census Bureau have redesigned the ASEC questionnaire for 2014, with Westat handling the redesign of the income questions, and the Census Bureau handling the redesign of the health coverage questions.

Three major structural changes were incorporated to address errors associated with the income questions:

1. Source-First Approach

Previous ASEC income questions have followed an interleafed design. That is, they ask about income source 1 followed by the amounts for income source 1; ask about income source 2 followed by the amounts for income source 2; and so on. This allowed respondents to become aware of the “consequence” of reporting income, and thus negatively affected reporting of income sources later in the interview, increasing under-reporting for those later income sources.

A source-first approach was therefore implemented for 2014. That is, respondents will first be asked about all income sources received, then a second pass will ask for amounts for those sources. In addition to better ensuring that all income sources have a fair chance of being reported, the source-first approach also gives respondents a second chance at thinking through an income source when amounts are requested, providing an opportunity for more specific recall, and perhaps triggering recall for other income sources.

2. Tailoring the order of income source presentation

For 2014, there are now three different orders in which income source questions may be asked:

• Default order (same as previous ASEC)

• Low-income order

• Senior (age 62+) order

The order presented will depend on details (such as roster and work history) collected from earlier in the interview, and from detailed information collected in prior interviews. The goal for using a tailored order is to reduce the effects of respondent fatigue by increasing the likelihood that the most relevant sources of income are discussed and identified earlier in the interview. The outcome is a reduction in missing and under-reported data.

3. Following “Don’t Know” Responses With an Amount Range

In previous years, when respondents did not know (or refused to tell) the value of their reported assets, interviewers simply moved on to the next asset. The redesigned income questions will now follow a “don’t know” or “refusal” response with a selection of ranges the respondent may choose from. For example, if asked how much Social Security was received in 2013, and the respondent replies “I don’t know”, we will now ask “Can you please tell me if you received less than $10,000, between $10,000 and $20,000, or over $20,000 for the total amount you received in Social Security Payments in 2013?”

The benefit to this approach is having an estimate of an amount rather than no answer at all, and may also help respondents develop a true point estimate of the amount.

More information regarding the changes for the income questions can be found in Attachment J.

The redesigned health insurance coverage questions address the following issues:

1. Recall Issue

Providing data on calendar year health insurance is an ASEC goal. However, a tendency of respondents is to ignore the calendar year reference period and instead report on their current status or their most recent spell of coverage. Respondents with more recent coverage are more likely to report accurate data than those with coverage in the distant past.

Therefore, the 2014 ASEC takes a new approach to questions on time period of coverage. First, current coverage status is determined, since this tends to be more accurately reported than past coverage. Then the duration of coverage is determined, at the month level, for the past calendar year. The new questions on current status may be leveraged to serve as an anchor which may help elicit reports of past year coverage more accurately than the standard methodology.

2. Respondent Fatigue

The household-level questionnaire design has shown to incur under-reporting for certain household members, but moving to a person-level design lengthens the survey.

A person/household level hybrid approach is now used. It begins by asking questions at the person level and if a particular plan type is identified, questions then determine if other members of the household are also covered by that same plan. For each subsequent person on the roster, if they have previously been identified as having a certain plan of coverage, that coverage is simply verified and they are asked if they had any additional plans. Persons not previously identified as having any coverage are asked the full set of coverage questions.

3. Comprehension Issue

Health plans were previously determined through a series of eight detailed questions on the source of coverage, which can challenge a respondent’s limited knowledge of the complexity of health insurance plans and programs.

To help reduce this type of error, the redesigned ASEC first asks about any coverage at all, then identifies a general source (e.g., job, government, or some other way), and then follows up with tailored questions to elicit the necessary detail.

More information regarding the changes for the health insurance coverage questions can be found in Attachment L.

Before implementing the changes to the ASEC, the Census Bureau conducted a 2013 ASEC Redesign Test. The primary objectives of the test are to evaluate whether changes to question wording, response categories, and redefinition of underlying constructs improve the quality of data collected. The initial stages of the Test consisted of content determination, cognitive laboratory pretesting, and expert reviews for the purpose of developing alternate versions of question content, which are described earlier within this Question. The field test portion was largely based on the data collection methodology changes suggested by Westat as presented in the attached report (Attachment J). The report justifies the changes made to ASEC questions, among them being to correct under-reporting of various sources of income, based on cognitive interviews they conducted.

The Redesign field test consisted of two panels – a control panel and test panel. The test panel consisted of retired CPS sample and received the redesigned ASEC instrument. The control panel received the traditional questions from the regular production ASEC.

The ASEC Redesign Test and the production ASEC differed in that the Test only used telephone interviewers from centralized telephone facilities; it did not include field interviewers nor did it include follow-ups for noninterviews. Attachment K describes the methodology, sample design, and research questions included in the Test, as well as decision criteria upon which final determinations were made regarding the questionnaire content.

Question 5. Contacts for Statistical Aspects and Data Collection

The following individuals may be contacted on the statistical, data collection, and analysis operations:

Statistical Collection Operations:

Lisa A. Clement

Survey Director, Current Population and American Time Use Surveys

Office of the Associate Director for Demographic Programs

U.S. Census Bureau

Washington, DC 20233-8400

(301) 763-3806

Survey Content:

Edward J. Welniak, Jr.

Social, Economic, and Household Statistics Division

U.S. Census Bureau

Washington, DC 20233-8500

(301) 763-5533

List of Attachments:

A - ASEC Questionnaire

B - Overview of CPS Sample Design and Methodology

C - Source and Accuracy Statement

D - CPS-263(MIS-1)(L), CPS Respondent Letter

E - Confidentiality Brochure

F - CPS-580(ASEC)(L), ASEC Respondent Letter

G - CPS-580(L)SP, ASEC Respondent Letter in Spanish

H - CPS-676, "Changing Situation" Pamphlet

I - CPS-676(SP), "Changing Situation" Pamphlet in Spanish

J - Cognitive Testing of Potential Changes to the ASEC

K - ASEC 2013 Redesign Test Results

L - Findings from a Field Test Experiment on a New Approach

M – Health Insurance in the Current Population Survey