0545 Addendum to ss 10-6-08

Addendum to Supporting Statement

0910-0545

Health and Diet Survey

Description of a Proposed Internet Survey to Collect Experimental Data Examining Differences between Phone- and Internet-Collected Data

The purpose of the study is to examine experimentally potential differences in fat knowledge information collected using Internet vs. telephone survey mode. The null hypothesis is different survey modes produce no response differences. The treatment group is Internet respondents drawn from an online consumer panel while the control group is a comparable size of telephone respondents obtained using list-assisted random-digit-dialing method.

FDA has been using telephone surveys (e.g., Health and Diet Survey, OMB #0910-0545) to ask consumers about knowledge of saturated fat, trans fat, and omega-3 fatty acids. With the growth in Internet use and response rate problems with telephone surveys, the Internet is increasingly used by various public and private organizations as a substitute for or a complement to telephone surveys when collecting consumer data. Nevertheless, there is still inconclusive evidence about whether Internet surveys provide information that is comparable to telephone surveys. Furthermore, the evidence may vary on a case-by-case basis. This study will attempt to establish some evidence on the comparability between telephone and Internet surveys for some key measures of interest to the FDA.

The Internet data will be used for methodological purposes rather than for policy purposes or for generating population representative estimates.

The focus of the proposed Internet survey will be two selected questions from the 2008 FDA Health and Diet Survey (HDS), which is approved by OMB and being administered by phone since September 2008. The two questions are: (Q3) awareness of three dietary fats (saturated fat, trans fat, and omega-3 fatty acids), and (Q4) their relationship with the risk of heart disease (see proposed questionnaire below). In HDS, these questions follow other questions such as (Q1) number of days in a week an adult eats breakfast, lunch, and dinner, and (Q2) exposure to nutrition information in restaurants. To minimize context effects, the proposed Internet survey will also begin by asking the latter two questions. Though we would prefer to include other HDS questions that precede the two dietary fat questions, our budget does not permit using that approach. We have made minor but necessary modifications to the wording of the HDS questions to facilitate their self-administration on the Internet.

As in the HDS, the universe of the proposed Internet survey is U.S. adults (18 years or older). Knowledge Networks (KN) will conduct the proposed Internet survey using its online consumer panel. Panelists will be invited to participate in this survey as one of the surveys they regularly receive from KN. Panelists will receive minor incentives to participate in surveys, generally equivalent to $1 but stated in the form of points. No survey specific incentive is anticipated. KN will draw a probability sample of 1,000 adults from the panel. The survey is expected to take 2 minutes to complete. There will be no overlap between the telephone and the Internet sample.

After the data are collected from the HDS and the proposed Internet survey, we plan to use regression analysis to examine the effects of mode of data collection (Internet vs. telephone) on fat knowledge responses. The analysis will control for demographic differences among respondents. We will compare the 1,000 respondents in the Internet survey with 100 randomly generated sets of respondents from the telephone survey, each set with 1,000 respondents. The target sample size of HDS is 3,000. The key measure of mode difference is the estimated coefficients of a zero-one dummy variable indicating mode.

We will be happy to share the results of the study with OMB.