NSC -- Sample change for Content Reinterview Study & Internet Mode Test and Addition of Internet Test Followup

OMB Non-substantive Change Request

Department: Commerce

Agency: U.S. Census Bureau

ICR Reference Number: 200910-0607-001

Title: American Community Survey Methods Panel Tests

OMB Control Number: 0607-0936

Expiration Date: 12/31/2012

I. Revised Timeframe and Burden Hours for the Content Reinterview Study

For the ACS Content Reinterview Study, the estimated completion time for an average household is 20 minutes per reinterview and the yearly sample is 72,000. Originally the burden was estimated assuming the program would run for 3 years; however, this study will only be in the field for 1 year. The burden for 3 years of this study was originally estimated at 72,000 hours and the removal of 2 years of the study will produce a 48,000 hour reduction to respondent burden.

Status	Cases	Burden Hours
Approved Burden	216,000	72,000
Revised Burden	72,000	24,000
Reduction in Burden	(144,000)	(48,000)

II. Revised PRCS Sample Size for Internet Mode Test

The initial projections were that the sample size for Puerto Rico would be 2,000 residential addresses (1,000 addresses each in two panels). Looking back at the high mail undeliverable rate in the PRCS coupled with the need to detect a meaningful difference on a fairly low mail response rate in Puerto Rico, we propose increasing the sample size to 7,400 addresses to increase our ability to detect a smaller difference between panels. The Internet Mode Test will consist of a national sample of residential addresses in Puerto Rico (3,700 addresses in each panel) with two experimental treatments with different Internet notification strategies:

• Not Prominent Offer - The Internet option will appear only on the questionnaire in a non-prominent place.

• Prominent Offer (Choice) - Households will be given a choice of completing the PRCS on paper or the Internet.  The Internet option will appear in the mail materials and be prominently displayed on the questionnaire.

The control panel for this test will consist of 3,000 PRCS production cases. The 7,400 sample size (3,700 cases per panel) will provide us with enough statistical power (.80) to detect a difference of about 3 percentage points in the overall response rates between the panels.

For an average household, the estimated time to complete the PRCS form is 38 minutes, including the time for reviewing the instructions and answers. The revised burden hours for a sample size with 7,400 residential addresses in Puerto Rico are 4,687 hours producing an increase of 3,520 burden hours.

Status	Addresses	Burden Hours
Approved Burden	2,000	1,267
Revised Burden	7,400	4,687
Increase in Burden	5,400	3,420

III. Additional Information on the 2011 ACS Internet Test Follow-up Interview

The clearance package submitted to OMB alluded to a reinterview for the Internet Test, but Census did not have a design at the time the clearance went through. We now have designed this interview. The follow-up interview will be conducted by CATI, and will include a sample of mail respondents, Internet respondents and nonrespondents. For Internet and mail responding households, the original respondent must be interviewed.

There are two components to the Internet Follow-up Interview.

1. Attitudes and Behavior Study (ABS) - these questions ask about Internet use/non-use, privacy concerns, and effectiveness of mailing materials. Note these questions will vary slightly across the three respondent groups: Internet respondents, mail respondents, and nonrespondents. (See Attachment A for the questions.)

2. Content Reinterview Assessment (CRA) - a subset of 14 ACS questions shown in the table in Attachment A will be re-asked using ACS production CATI wording to look at consistency of responses. The questions were selected to represent various Internet question format types. Additionally, questions were chosen that would be less likely to have recall bias and time specific responses such as “during the past week.” (See Attachment B for a list of the questions.)

The ABS questions will be asked first. Note that only a sample of Internet and mail respondents will get this portion of the interview while this is the only part of the follow-up interview that nonrespondents get. CRA questions will be asked of Internet and mail responding households only for which the original interview was classified as a sufficient partial or a completed interview. CRA person level questions will be asked about all household members from the original interview roster.

For the CRA, we will compare gross difference rates for Internet and mail respondents. The gross difference rate measures the percentage of inconsistent responses between the original interview and the reinterview. Using an 80% power analysis, we determined that we need to sample 6,000 households with phone numbers for each mode in order to detect a difference in gross difference rates of 4.0 percentage points or higher between modes using a significance level of α = 0.10. This sample size was determined assuming an 85% response rate to the reinterview and an 85% success rate interviewing the same respondent.

The ABS analysis is a qualitative analysis that needs minimal sample to make conclusions. The sample sizes for respondent groups are shown below:

1. Internet respondents: Since we need 400 complete ABS interviews (100 per Internet test panel (i.e., notification strategy) excluding the control), we will subsample 550 households for ABS interview. This subsampling accounts for our 85% success rate in our content tests in reaching the same respondent and an 85% response rate (as seen in previous content tests).

2. Mail respondents: Since we need 500 complete ABS interviews (100 per panel including control) with the same respondent and based on our success at reinterviewing these households in past content tests (85% response rate and 85% same respondent rate), we would need 700 households to be subsampled for the ABS interview.

3. Nonrespondents: We need 400 complete interviews (100 per panel excluding the control panel). Taking into account our estimated telephone look-up success (55%), expected response rate for a difficult population (30%), and doing an adjustment (85%) for reaching the person who typically handles the mail (the person needed for the interview), we would need roughly 2,900 nonresponding households sampled.

Total sample for the follow-up interview is 14,900 (6,000 for Internet respondents, 6,000 for mail respondents, and 2,900 for nonrespondents). For an average household, the estimated time to complete only one component of the follow-up interview is 10 minutes. The estimated time to complete both components of the interview is 15 minutes. Only 1,250 households will be asked both components for a total burden of 312.5 hours. The remainder of the sample (13,650 households) will only be asked one component for a total burden of 2,275 hours. Follow-up interviews will begin week of March 14, 2011 and continue to April 18, 2011.

Group	Cases	Burden Hours
Both follow-up components	1,250	312.5
Only one component	13,650	2,275
TOTAL		2,587.5