(posted 10/05/2006)

Statement of the Problem:

In the 2001 SIPP Panel, both parents (or guardians) were asked questions about receiving Federal Supplemental Security Income (SSI), and Social Security Administration (SSA) payments for themselves, and on behalf of their dependent children. In the 2004 Panel, in an effort to reduce respondent burden, SIPP only asked these questions of one parent (or guardian). In processing the 2004 SIPP data, the edits were looking for an answer from both parents (or guardians), but since these question were only asked of one parent, the information the second parent (or guardian) was imputed. This resulted in high allocation rates, as well as an higher response rate for the variables ESSCHILD, ESSICHLD.

Resolution of the Problem:

This inconsistency in the edit was realized and corrected for Wave 2 and forward. The edit was changed to copy the answer from the guardian's record onto the other parent's record, thus reducing the allocation rates to an acceptable level, and made the data more consistent to the SIPP 2001 Panel data. This change to the edit was not made in time for the Wave 1 processing and data users should be aware of the issue when using these variables for Wave 1.

Wave 1 of the SIPP 2004 Panel has been re-released. In the latest release the inconsistency between Panels has been applied to Wave 1. All allocation rates are now consistent with the 2001 Panel.

(posted 01/08/2007)

Statement of the Problem:

In the 2001 Panel the lump sum questions were asked in the General Income section of the instrument. In the 2004 Panel these question were moved to the Labor Force section although the follow up monthly recipiency and amounts were still asked in the General Income section.When a respondent answered that they received a lump sum in Labor Force, the monthly recipiency and amount questions of the General Income section were not being set in the instrument. This caused high imputation rates for GICODES 15, 39, and 52.

Resolution of the Problem:

This error in the instrument was not detected in time to make a correction to the instrument. Data user's should just be aware of these high imputations.

(posted 05/11/2007)

Statement of the Problem:

There was an error in the edit program that affected variable EWRKEXP3, and caused high imputation rates. This questions asks if the state or county welfare office required participation in the specific work experience. There was a skip error which led some of the given universe to skip out of the series of work experience questions before answering EWRKEXP3.

Resolution of the Problem:

This error in the edit program was not detected and fix until Wave 3+. Data user's should just be aware of these high imputations for Waves 1 and 2.

(posted 05/11/2007)

Statement of the problem: 

Starting in wave 2 of the 2004 panel the SIPP instrument did not capture all potential in universe respondents for questions regarding reasons for transitioning from non-participation to participation in the first month of the reference period. The instrument only collected data for individuals new to the survey that reported receiving cash or in-kind assistance in the first month of the reference period. The universe for these questions should have also included prior wave respondents in the survey who indicated they had not participated in a program in the last month of the prior wave, but indicated they participated in a program in the 1st month of the current wave (for example, a respondent who did not report receiving TANF in month 4 of wave 2 and who did report receiving TANF in month 1 of wave 3).

The questions involved are:

RAB1R1: First reason applied for Pub Asst/AFDC the 1st time.
RAB1R2: Second reason applied for Pub Asst/AFDC the 1st time.
RAB2R1: First reason applied for Pub Asst/AFDC the 2nd time.
RAB2R2: Second reason applied for Pub Asst/AFDC the 2nd time.

RFB1R1: First reason applied for food stamps the 1st time.
RFB1R2: Second reason applied for food stamps the 1st time.
RFB2R1: First reason applied for food stamps the 2nd time.
RFB2R2: Second reason applied for food stamps the 2nd time.

RWB1R1: First reason applied for WIC the 1st time.
RWB1R2: Second reason applied for WIC the 1st time.
RWB2R1: First reason applied for WIC the 2nd time.
RWB2R2: Second reason applied for WIC the 2nd time.

RGB1R1: First reason applied for General Assistance the 1st time.
RGB1R2: Second reason applied for General Assistance the 1st time.
RGB2R1: First reason applied for General Assistance the 2nd time.
RGB2R2: Second reason applied for General Assistance the 2nd time.

RSB1R1: First reason applied for SSI the 1st time.
RSB1R2: Second reason applied for SSI the 1st time.
RSB2R1: First reason applied for SSI the 2nd time.
RSB2R2: Second reason applied for SSI the 2nd time.

RSB1R1: First reason applied for other welfare the 1st time.
RSB1R2: Second reason applied for other welfare the 1st time.
RSB2R1: First reason applied for other welfare the 2nd time.
RSB2R2: Second reason applied for other welfare the 2nd time. 

(posted 06/25/2007)

Statement of the Problem:

There was an error starting in Wave 3 for the variable EVETTYP( what type of payment does the respondent receive from the VA). This question had an error in the instrument that caused the prior wave instrument data to be stored incorrectly. The post-collection edit program retrieves the correct data by using the edited previous wave information, but sets an allocation flag of a 4, indicating that the current wave data is based upon what was reported in the prior wave.

(posted 02/08/2008)

Statement of the problem: 

In a change from previous panels, beginning in wave 1 of the 2004 panel, the amounts in the variables TPRFTB1 and TPRFTB2 (net profit or loss from a business) were intended to be apportioned by month to the monthly amounts within each set of the following recodes: THEARN , THTOTINC, TFEARN, TFTOTINC, TSFEARN, TSTOTINC, TPEARN, and TPTOTINC. Through an error, this change was not made on the previous releases of the public use files for Waves 1 to 4 of the 2004 panel. These files have been re-run to correct the problem, and re-released on 01/29/08. In previous panels, some or all of the corresponding profit from a business may have been reported as business income (TBMSUM1 — TBMSUM4), and would therefore have been included in these recodes as part of their business-income component. The degree to which the direct and complete addition of business profit to the recodes in 2004 affects their practical comparability with the data from previous panels is unknown.

Resolution of the Problem:

The public use files for Waves 1–4 have been re-run to correct this error. The re-release date was 01/29/08.

(posted 02/08/2008)

Statement of the Problem:

It was discovered in post data processing of the SIPP 2004 Panel that there was an error in the edit program for how amounts were imputed for the General Income section. The edit program was imputing erroneous amount values for respondents that did not record an amount for the current wave, and did not have amounts from the previous wave. These cases were imputed from a deck that stored amounts from all general income codes. This allowed high values to be stored in programs that usually do not have such amounts.

Resolution of the Problem:

This error in the edit program was corrected starting in Wave 5. The edit was changed so that these cases were imputed from the current wave hotdeck. Data users should be aware that these extreme outliers are possible for waves 2-4.

(posted 04/07/2008)

Statement of the Problem: 

An error was discovered in the technical specifications in the General Income section for Waves 1-5 on the public use file. The variables T60AMTT and T60AMTG were improperly edited. This caused a decrease in the number of people in universe for these items and invalid data to be output. This data should not be used.

Resolution of the Problem:

Starting in Wave 6 this error was corrected and the variables are showing valid data.

(posted 04/07/2008)

Statement of the Problem: 

The variables listed below, which focus on cash and in-kind assistance from a variety of government and non-governmental sources, are asked of SIPP respondents regarding the 4-month reference period. For example, if the reference period for data collection were the 4-month period January through April, the question associated with EFOODTP1 would be:

Earlier you said that since January 1st (the beginning of the reference period) you received food assistance. Did you receive money, vouchers, or certificates to buy groceries or food? 
1 = Yes
2 = No

The public use file is output in a person-month format (each person has monthly record for the months they are in sample in the reference period and each record has a monthly variable assigned). The creation of this file transforms these variables from a wavely-level to a monthly-level for consistency. Users need to be aware that in doing longitudinal analysis the variables listed below should not be viewed as having a monthly value.

For example, over the course of a wave, the variable EFOODTP1 will have 4 values, one for each month of that wave. If a respondent reports “yes” for that wave all of the corresponding monthly values are “yes.” However, that does not mean that the respondent received money, vouchers, or certificates in each month. The analysis is limited to the fact that over the 4-month period the respondent received this type of assistance at some point in time – it is not known in how many months assistance was received.

Because of this, only the first monthly value for the 3 waves that make up the calendar year should be used in the longitudinal analysis, since they reflect the data for the entire reference period.

The following list of variables in the 2004 SIPP panel are collected for the wave and output at the monthly level.

EPAOTHR1 APAOTHR1 EPAOTHR2 APAOTHR2 EPAOTHR3 APAOTHR3
EPAOTHR4 APAOTHR4 EPAOTHR5 APAOTHR5 EPAOTHR6 APAOTHR6 
ETRANTP1 ETRANTP2 ETRANTP3 ETRANTP4 ETRANTP5 ATRANTP
EGASSCE1 EGASSCE2 AGASSCE 
ETOKSCE1 ETOKSCE2 ATOKSCE 
EFOODTP1 EFOODTP2 EFOODTP3 EFOODTP4 AFOODTYP 
EFOODSC1 EFOODSC2 EFOODSC3 EFOODSC4 AFOODSCE
ECLOTHTP ACLOTHTP
ECLTHSC1 ECLTHSC2 ECLTHSC3 ECLTHSC4 ECLTHSC5 ACLTHSC 
EPUBHSTP APUBHSTP 
EPUBHSC1 EPUBHSC2 EPUBHSC3 EPUBHSC4 APUBHSC 
ECASHSC1 ECASHSC2 ECASHSC3 ECASHSC4 ACASHSCE 
ECASHGVT ACASHGVT 
EPACASH1 APACASH1 EPACASH2 APACASH2 EPACASH3 APACASH3
RAB1R1 RAB1R2 RAB2R1 RAB2R2 RAS1 RAS2 RWB1R1 RWB1R2 RWB2R1
RWB2R2 RWS1 RWS2 RFB1R1 RFB1R2 RFB2R1 RFB2R2 RFS1 RFS2 RGB1R1
RGB1R2 RGB2R1 RGB2R2 RGS1 RGS2 ROB1R1 ROB1R2 ROB2R1 ROB2R2 ROS1
ROS2 RSB1R1 RSB1R2 RSB2R1 RSB2R2 RSS1 RSS2

(posted 07/29/2010)

Statement of the Problem: 

Reprocessing 2004 SIPP, waves 1-12 to correct for instrument and processing errors to the monthly amounts of social security benefits received:

An error was discovered in the way that the 2004 SIPP collected and processed the monthly amount of Social Security benefits an individual received for retirement or disability. Starting with the beginning of the 2004 panel, the instrument attempted to collect the amount of Social Security income that was withheld for the Medicare Part B premium. The instrument, however, did not correctly capture whether the Social Security benefits that were reported included or excluded the monthly Part B premium. For waves 1-4, the universe was too large, and was asked of all individuals, age 18+. Starting in waves 5-12, the universe for the question dealing with the inclusion or exclusion of the Medicare Part B premium from the reported monthly Social Security benefits was unintentionally restricted to individuals age 65+ and (rather than or) having a disability. This universe restriction was still in place at the beginning of the 2008 SIPP panel, but has since been corrected in the data collection instrument.

In addition, several issues were discovered regarding how the processing system was handling the collected data. In the processing system the instrument variable that identified the inclusion or exclusion of the Medicare Part B premium (USSPRTB) from the reported amount of Social Security benefits received was not being properly picked up, in that no universe check was being applied to determine if the response to USSPRTB should have been used to adjust the amount of Social Security benefits received. In addition, for those individuals that reported receiving Social Security benefits jointly with their spouse the processing system was not distributing the amounts between the two individuals, but was assigning both individuals a monthly amount of $1.00. And, starting in wave 2 and beyond, when feedback data from the previous wave was used to record the current wave amount of Social Security income, the processing system was automatically adding the Medicare Part B premium amount to the prior wave amount that was being brought forward.

Variables directly impacted:

T01AMTA – the monthly amount of Social Security income received by an individual
TFSOCSEC – the total monthly amount of Social Security income received by a family
THSOCSEC – the total monthly amount of Social Security income received by a household
TSSOCSEC – total monthly amount of Social Security income received by a related subfamily.

The instrument and processing errors did not impact the estimates of the number of people, families and households that reported receiving Social Security benefits in the 2004 SIPP panel. Rather, the errors in the monthly amount of benefits received tended in some cases to over-estimate benefits for individuals, families, and households, while under-estimating these benefit amounts in other cases. Therefore, in the aggregate, the total person monthly income amounts, total family monthly income, and the total household monthly income amounts reported in the 2004 SIPP panel, as well as the respective allocation rates, were comparable to those aggregate monthly amounts and allocation rates found in the 2001 SIPP panel.

In response to these errors, it is suggested that SIPP users of the data in the 2004 panel need to be careful on the interpretation of their research findings when their primary interest  is the amount(s) of Social Security benefits received, over the course of the 2004 panel, for individuals, families, and households that reported receiving retirement or disability benefits.

Resolution of the Problem:

All waves in the 2004 SIPP panel have been reprocessed to correct for the instrument and coding errors found in the processing system regarding the amount of monthly social security benefits received.  In reprocessing each wave of the 2004 panel, a rule was implemented that randomly allocated a fixed dollar amount for the Medicare Part B premium to approximately 80 percent of respondents who reported receiving a monthly social security benefits because of retirement or a work-ending disability.  Users of this data need to be aware that the allocation rule was implemented for each wave, independent of the prior wave response, which resulted in some individuals being allocated a Part B Premium in one wave and, potentially not being allocated a premium amount in a preceding wave(s).  Thus, over the 48 month duration of the 2004 SIPP panel monthly social security amounts for some individuals, families, and households will experience fluctuations by the fixed dollar amount of the Medicare Part B premium implemented in by the reprocessing rule.

Weighting Changes:

Some of the variables impacted by the reprocessing of waves 2+ in the 2004 SIPP panel were also used in the non-interview adjustment procedure.  This resulted in changes to the non-interview factors, resulting in changes to the final monthly weights associated with most respondents.  However, the overall impact on change on the magnitude of the monthly weights was minimal.  For example, in wave 2 the magnitude of the change in the monthly weight for 87 percent of the cases was 1 or less and for 99 percent of the cases the magnitude of the change in the monthly weight was 5 or less.