Questions and Answers on the SIPP 2008 Panel

Questions and Answers on the SIPP 2008 Panel OMB Submission

1. In section A.16 you noted that you will release cross-sectional, topical module, and longitudinal reports, but only mentioned releasing cross-sectional and topical module data files. Do you also plan to issue longitudinal data files or would users match the cross-sectional files themselves to create longitudinal files or ?

Starting with the 2004 Panel, we are not longitudinally editing the data. The data users must match the cross-sectional files to create a longitudinal data file. The Census Bureau releases SIPP reports dealing with longitudinal issues from an internal matched data file.

2. B.1 notes that you are over sampling low income households, but it’s not clear how you are doing this. Please provide more information.

Generally, we have about an 11% increase in the number of low income households

(< 150% poverty) with the oversample. In terms of how it's done, addresses are classified into 2 strata and those in the low income strata are sampled at a slightly higher rate than the other stratum. Census 2000 long form income data is used to determine whether the household at the address has income < 150% of the poverty threshold determined for that household. If income data is not available for a particular address, then the block group estimate of the percent of households < 150% poverty is used to assign an address to the appropriate stratum.

There is no direct oversample of minorities, but the oversample of low income does increase our representation of Blacks and Hispanics. Looking at Wave 1 data from the 2004 Panel we have about 13% more Blacks and 11% more Hispanics in sample than we would have had with a self-weighting design.

3. Please provide more information on expected response rates for later waves, including actual response rates from the completed 2004 waves, and your plans to examine potential non-response bias due to attrition.

We expect the response rates for the 2008 Panel to be similar to the rates from the 2004 Panel. Attached is the Sample Loss Report through Wave 11 for the 2004 Panel.

Our plans for examining non-response bias due to attrition include a comparison of frame characteristics for Wave 1 nonrespondents to Wave 1 respondents, calculating weighted response rates for different subgroups of interest using Wave 1 characteristics, and beginning with Wave 2, we will compare current wave nonrespondents to respondents using Wave 1 characteristics.

4. I would like to obtain more information about the use of incentives by the FRs. Specifically, what parameters or instructions are interviewers given for using incentives? Also, what percentage of cases received incentives in the recent waves of the 2004 panel (please note that if this information will take time to gather, you could report back to me later on this; if it isn’t available at all, I would like it to be tracked in the first couple of waves of this panel and reported upon completion)?

In the 2004 Panel, a $40 debit card incentive was offered to approximately 20 percent of the sample units if they refuse or potentially refuse participation in any wave of the survey. All sample cases are eligible to receive an incentive during the life of the panel. The (S)FRs have the discretion to determine which cases will receive the incentive, however, the incentive is conditional on the household responding to the survey. In order for a household to receive an incentive, they must complete a SIPP interview.

We are putting together information regarding the number of cases that received an incentive.

5. The FAQs note that the census Bureau will obtain information about program participation from other government agencies, but allows the respondent to opt out by telling the field representative at the time of the interview. Since this sheet appears to be included with the advance letter, is this topic mentioned by the FR at any point during the interview? Where does the FR indicate if the respondent asks that their information not be combined with information from other agencies?

The opt out topic is not specifically mentioned by the FR; however, if the respondent indicates they do not want the Census Bureau to combine their SIPP data with information from other agencies, the FR has a function key in the CAPI instrument to capture the respondent identification information for all of the persons in the household that do not want their information combined. A flag is set for each person that doesn’t want their information combined with data from other agencies.