This User Note contains information for several health insurance variables applicable to the 1996 SIPP Panel. The Wave(s) each 1996 Panel problem applies to are specified below. The problems will be corrected in the 1996 Longitudinal file and in future (2001 and beyond) panels.

In Wave 1 of the 1996 Panel, during the postdata collection process, only categories 1 - 7 were considered valid answers to EHEMPLY. All of the 1,024 responses equal to a value of "8" (Other) were imputed and distributed over categories 1 - 7.

The Wave 1 public use documentation for EHEMPLY incorrectly reads as follows:

"What was the source of ...'s health insurance?"

-1 Not in universe
1 Current employer or work
2 Former employer
3 Union
4 CHAMPUS
5 CHAMPVA
6 Military/VA health care
7 Other

For variable EHEMPLY (source of ... 's health insurance) in Wave 1 - Wave 12, the instrument accepted as valid answer categories 1 - 8 as listed below:

"What was the source of ...'s health insurance?"

-1 Not in universe<
1 Current employer or work
2 Former employer
3 Union
4 CHAMPUS
5 CHAMPVA
6 Military/VA health care
7 Privately purchased
8 Other

For Wave 1, answer category "7" should read "Privately Purchased" not "Other."

This problem was corrected in both the postdata edit processing, and the data dictionary public use documentation for Wave 2 - Wave 12.

The universe statement for EHEMPLY is incorrect. In the 1996 SIPP Panel Waves 1 - 12 the edited data allows some respondents with EHIOWNER = 2 (Covered by someone else's plan) to be asked EHEMPLY when the holder of the health insurance lives outside of the household. That is, when ENONHH = 1. The existing universe statement excludes these people.

EHIOWNER asks, "Was the coverage in ...'s own name or was ... covered as a family member on someone else's plan, both or neither?"

1 Covered in own name
2 Covered by someone else's plan
3 Covered both in own name and by someone else's plan
4 Not covered<

EHEMPLY asks, "What was the source of ...'s health insurance?"

-1 Not in universe
1 Current employer or work
2 Former employer
3 Union
4 CHAMPUS
5 CHAMPVA
6 Military/VA health care
7 Privately purchased
8 Other

The universe for EHEMPLY is listed in the metadata as: EPOPSTAT =1 and EHIOWNER = 1 or 3.

The correct EHEMPLY universe is: EPOPSTAT = 1, EHIOWNER = 1 or 3 and (EHIOWNER = 2 where ENONHH = 1).

That is, if a respondent reports they are covered as a family member on someone else's plan (EHIOWNER = 2) and the family member who's plan they are covered under does not live in the household (ENONHH = 1), then the respondent is asked the source of the health insurance (EHEMPLY).

There are cases where a respondent reported coverage by someone else (EHIOWNER = 2), but were not reported as being covered by anyone else in the household.

In these instances, ENONHH is set to "1" (Covered by someone outside the household.) This puts those people (who reported coverage by someone else) in the universe for an answer to EHEMPLY. An answer for EHEMPLY is then imputed for such respondents.

Variable EHIMTH - the question incorrectly includes "military related health care" in the Wave 1 - Wave 12 public use documentation data.

EHIMTH currently reads as follows:

Was ... covered by a health insurance plan other than medicare, medicaid, or military related health care in this month?

1 Yes, covered 
2 No, not covered

EHIMTH should read:

Was ... covered by a health insurance plan other than medicare or medicaid in this month?

1 Yes, covered 
2 No, not covered

This is an error in the public use documentation only. "Military related health care" was not included in this question in the instrument.

Additional information:

EHIMTH - There are a few cases on the file where a person reports ownership of a health insurance policy but reports "Don't know" or "Refused" in the follow up question asking who is covered by that policy. The program as it was run for the 1996 Panel gives coverage to all spouses and children of this individual in coverage units for all months. However, EHIMTH is set to "1" for these people only for the first reference month. This causes inconsistencies between reported coverage in EHIMTH and RCUTYP58.

The universe statement for RPRVHI is incorrect. This recode is created for persons covered by a health insurance policy provided through an employer (current or former), union, or "privately purchased" source at any time during the reference period. The universe currently reads "All persons covered by health insurance (EHIMTH = 1)."

The universe statement should read: "All persons covered by private health insurance at any time during the reference period. This does not include children covered by health insurance from someone outside the household or from an undisclosed source. Unit owner has EHEMPLY = 1, 2, 3 or 7."

General statement: There are inconsistencies between the variables EHIMTH and RCUTYP58 identified below in Parts 1, 2 and 3.

There are a few cases on the file where the edited value for RCUTYP58 = 1, "Covered by health insurance" and the value for EHIMTH = 2, "Not covered". These cases occur when a person reports ownership of a health insurance policy, but reports "Don't know" or "Refused" in the follow up question asking who is covered by that policy. The program as it was run for the 1996 Panel gives coverage to the spouse and children of this individual in coverage units (RCUTYP58) for all months. However, EHIMTH is set to "1" for these people only for the first reference month. This causes inconsistencies between reported coverage in EHIMTH and RCUTYP58. The value of EHIOWNER is not reset to "2" (Covered by someone else's plan) for this same universe of people who are given coverage only in month 1. They have a value of "4" (Not covered) which is inconsistent.

The value of EHIMTH should be set to 1 for the remaining months of the reference period. EHIOWNER should be set to "2" (Covered by someone else's plan.)

In addition, a relatively small number of cases in Waves 2 through 5 show inconsistent answers between these two variables when the value of EHIOWNER is "1." In these cases, EHIMTH=1 and RCUTYP58<>1. The inconsistencies were caused by an indexing error which was fixed in Wave 6.

In all of the cases mentioned in Problem 5 Part 1 and 2, the individuals were covered by health insurance in the given month and RCUTYP58 should have been equal to "1."

The vast majority of inconsistencies between EHIMTH and RCUTYP58, when EHIMTH=1 and RCUTYP58 = 2 occur when an individual lists the owner of his health insurance as someone other than himself (EHIOWNER = 2). The recode RCUTYP58 is set based on coverage units. A coverage unit for health insurance contains the owner of the unit, (EHIOWNER = 1 or EHIOWNER = 3) and the persons in the household that the owner reports covering. If the owner does not report covering an individual, that individual is not included in the unit. The majority of the inconsistencies are caused because no consistency check is done in cases where EHIOWNER = 2 to verify that the individual is indeed covered by someone else . If a person reports coverage, but says he/she is covered by someone else, or does not report the owner, and EHIOWNER is imputed as a "2," this person only receives a value of "1" in RCUTYP58 if someone else in the household reports covering him/her. These cases include adults who report coverage from someone outside the household.

In these cases, the reported coverage in EHIMTH is correct and the value in RCUTYP58 can be changed to "1". However, we do not know the owner of the insurance policy covering these people. In future Panels these individuals will be assigned coverage from someone outside the household (ENONHH = 1).

There are instances of inconsistencies between the values of RCUTYP57 and ECDMTH. For example, there are cases where RCUTYP57 = 2, and ECDMTH = 1, and vice versa.

There should be no difference between the values of these two variables. This problem occurred in Waves 1 - 5, and was resolved from Wave 6 on. Data users can safely use ECDMTH as the correct answer for Medicaid monthly coverage status for Waves 1 - 5.

There is a large number of children (for example 1500+ persons/6000+ monthly records in Wave 2) who are not reported as being covered by an adult in the household but who are reported as having health insurance coverage in screens H1KDCOV or H2KDCOV. These children are made owners of their own health insurance coverage units (RCUOW58A is set to the child's person number). This appears to be done correctly. The follow-up questions (HI1OUT and HI2OUT) ask, for single child and multiple children households, which of these children are covered by someone outside the household. The code for multiple children households works correctly and sets EHIOWNER equal to "2" (Covered by someone else's plan) for those children reportedly covered by someone outside the household. For those not reported, the value of EHIOWNER is set to "1" (Covered in own name.) However, the children in single child households are not reset to "2" (Covered by someone else's plan) where appropriate.

There is no resolution for these cases in the 1996 Panel cross-sectional data. However, this error will be corrected in the release of the longitudinally edited data.

The recode ENONHH is incorrect. ENONHH asks, "Was ... covered by a health insurance plan of someone who does not currently live in the household?" The universe for this variable is all persons. The adults who reported coverage from outside the household in screen HIHOLDR reflect the correct value in the recode, however, no children are reported in the recode as having coverage from outside. For example, in the Wave 2 unedited variables, there are 389 children in single child households who reported coverage from outside the household (screen HI1OUT), and 533 kids in multiple child households who reported coverage from outside (screen HI2OUT). There should have been 922 children who had ENONHH set to a "1" (yes) however they remained set to a "2" (no)."

A partial resolution to this problem is to set ENONHH to "1" for children who report themselves as owner of their health insurance unit (EPOPSTAT = 2 and RCUOW58A = EPPNUM) but EHIOWNER = 2. Because of the error reported in Problem 7 above, this solution does not work for children in single child families as they do not have EHIOWNER set to "2".