Using CPS Health Insurance Coverage Variables

(updated 1/25/06)

(Printing note: use landscape format)

The March CPS has a wealth of information regarding a person's health insurance coverage. However, this abundance of information can make the task of using the variables somewhat daunting. This section attempts to give some order to these variables. To begin with, note that there are two groups of coverage variables -- those pertaining to "current" coverage (i.e. as of the time of the survey) and those pertaining to "previous year" coverage (i.e. covergae any time during the previous year, but not necessarily the whole year). For most applications, TRIM uses the latter group of variables, so the remainder of this document focuses only on those. Note that this information applies only to CPS variables and new composite variables constructed directly from the CPS variables. For information on TRIM-simulated health insurance coverage variables, see the descriptions of the various health modules.

The following table lists the CPS variables most commonly used to describe coverage. For each of these variables, the table shows the corresponding "detailed" TRIM input variable. There is a one-to-one correspondence between these detailed TRIM input variables and the CPS variable. However, TRIM also creates some "aggregate" coverage input variables by combining certain information from the detailed variables. Usually, this means combining the information given by the catch-all "other type" variables (OTHSTYP1-6 and AHITYP1-6) with the variable that specifically asks about that type of coverage. Note that for all except the catch-all variables, a value of "1" means the person is covered by that type of insurance.

IMPORTANT NOTE REGARDING THE CATCH-ALL VARIABLES:
OTHSTYP1-6 variables have been available since the 1995 March CPS. Beginning with the 2001 March CPS, Census added another catch-all question to the survey and, as a result, additional catch-all health insurance coverage fields, AHITYP1-6. The coding of each set of catch-all variables is the same and is shown below for the most recent data years. Prior to the 2001 survey, the range was 1-14, and CHIP coverage was not separately reported. However, beginning with the 2001 survey, Census inserted the CHIP code in the middle of the coding scheme, as shown below. The TRIM3 names for these fields are HealthCoveredOtherType1-6 (for CPS variables OTHSTYP1-6) and OtherHealthInsuranceCoverage1-6 (for CPS variables AHITYP1-6).

CPS Name	Position	CPS Variable Description	Detailed TRIM input variable	Aggregate TRIM input variable
HI	748	covered by current or former employer or union (exclude military)	HealthCoveredByOwnPlan	HealthEmployerUnionPlan
DEPHI	750	covered by employer or union plan (dependent)	HealthCoveredDependent	HealthCoveredDepGroupPlan
PRIV	757	covered by plan purchased directly	HealthCoveredOwnPrivate	HealthCoveredOwnNonGroup
DEPRIV	759	covered by plan purchased directly (dependent)	HealthCoveredDependentPrivate	HealthCoveredDependentNonGroup
OUT	765	covered by plan of someone outside the household	HealthInsuranceOutsideHh	HealthCoveredOutsideHh
CARE	766	covered by Medicare	HealthCoveredMedicare	HealthMedicareCoverage
CAID	767	covered by Medicaid	HealthCoveredMedicaid	HealthMedicaidCoverage
OTH	770	covered by other including Champus, ChampVA, VA, military, Indian (see OTYP-1 - OTYP-5 for type)	HealthCoveredOtherGov
OTYP-1	771	covered by Champus	HealthCoveredChampus	HealthChampusCoverage
OTYP-2	772	covered by ChampVA	HealthCoveredChampVA	HealthChampusCoverage
OTYP-3	773	covered by VA or military health care	HealthCoveredVA	HealthChampusCoverage
OTYP-4	774	covered by Indian Health	HealthInsuranceIndian	HealthCoveredIndian
OTYP-5	775	covered by other	HealthCoveredOther	HealthCoveredOtherNonGov
OTHSTPER	776	covered by other type of health insurance (including a state plan)	HealthCoveredOtherState
PCHIP	880	child under 19 and with no medicaid coverage covered by SCHIP	CHIPCoverage	HealthCHIPCoverage
"Catch-all" questions:
OTHSTYP1-	777-	other type:	HealthCoveredOtherType1-
OTHSTYP6	787	1=Medicare	HealthCoveredOtherType6	HealthMedicareCoverage
AHITYP1-	867-	2=Medicaid	OtherHealthInsuranceCoverage1-	HealthMedicaidCoverage
AHITYP6	877	3=CHAMPUS	OtherHealthInsuranceCoverage6	HealthChampusCoverage
		4=CHAMPVA		HealthChampusCoverage
		5=VA health care		HealthChampusCoverage
		6 = Military Health Care		HealthChampusCoverage
		7 = CHIP		HealthCHIPCoverage
		8 = Indian Health Service		HealthCoveredIndian
		9 = Other government		HealthChampusCoverage
		10 = Employer/Union (Policyholder)		HealthEmployerUnionPlan (if age≥15), HealthCoveredDepGroupPlan (if age<15)
		11 = Employer/Union (Dependent)		HealthCoveredDepGroupPlan
		12=Private (Policyholder)		HealthCoveredOwnNonGroup (if age ≥15), HealthCoveredDependentNonGroup (if age <15)
		13 = Private (Dependent)		HealthCoveredDependentNonGroup
		14 = Outside household		HealthCoveredOutsideHh
		15 = Other		HealthCoveredOtherNonGov

There are three additional TRIM aggregate coverage input variables which summarize non-government coverage at an even higher level. They are "HealthGroupCoverage", "HealthCoveredNonGroup", and "HealthCoveredByOtherPlan". These higher-level aggregate variable indicate coverage if certain lower-level aggregate variables indicate coverage as follows:

HealthGroupCoverage will indicate coverage if HealthEmployerUnionPlan or HealthCoveredDepGroupPlan indicate coverage.

HealthCoveredNonGroup will indicate coverage if HealthCoveredOwnNonGroup or HealthCoveredDependentNonGroup indicate coverage.

HealthCoveredByOtherPlan will indicate coverage if either of the previous two higher-level aggreagate variable indicate coverage, or if HealthCoveredOutsideHh indicates coverage.

With this information, we can see how the aggregate TRIM coverage input variables can be used to describe a person's health insurance coverage. Most broadly, a person's coverage can be broken down into government and non-government. If HealthCoveredByOtherPlan indicates coverage, the person is covered by a non-government program. If HealthMedicareCoverage, HealthMedicaidCoverage, HealthChampusCoverage, HealthCoveredIndian¹, HealthCoveredOtherNonGov², or HealthCHIPCoverage indicate coverage, the person is covered by a government program. Within the non-government coverage category, a person's coverage can be broken down into group and non-group. If HealthGroupCoverage or HealthCoveredOutsideHh³ indicate coverage, the person is covered by a group plan. If HealthCoveredNonGroup indicates coverage, the person is covered by a non-group plan. Keep in mind that these variables indicate coverage as reported by the CPS. For some analysis, it may be preferable to use the TRIM-simulated coverage variables.

The Timing of Coverage

When discussing the concept of "coverage", one must be clear as to what the applicable time frame is. The survey questions which generate the coverage variables discussed in the previous section are intended to capture coverage sometime during the previous year, but not necessarily the whole year. This is an "ever-on" concept of coverage. It is important to note that at no point are respondents asked if any members of the household were ever uninsured for all or part of the previous year. Consequently, estimates of the uninsured are estimates of persons uninsured for the entire year. However, it should also be noted that some studies suggest that the responses to these questions are actually a mix of point-in-time responses and ever-on responses⁴.

For some analyses a "point-in-time" concept is desirable. Such a concept is comparable to taking a "snapshot" of people's actual coverage status at one point of the year. As mentioned in the previous section, the coverage variables on the CPS are divided into two groups -- those pertaining to "previous year" coverage, and those pertaining to "current" (i.e. the time of the survey) coverage. While in theory the latter group could be used to estimate "point-in-time" coverage, in fact they were found to be unreliable, and were soon discontinued as part of the CPS survey.

Somewhere between these two extremes is the "average monthly" or "full year equivalent" concept. This concept tries to adjust the "point-in-time" estimates by averaging them across the year, thus giving an indication of what coverage is like during an "average" month. Among the available coverage variables in the TRIM system, only simulated Medicaid coverage provides users with the data for computing average monthly values (see the documentation for the Medicaid module for details). For other types of coverage, assumptions would have to be made to allocate the annual "ever-on" coverage across months. For example, for employer-based coverage types, the assumption could be made that it only applies to months when the person is working. However, such an assumption may not be true, so it is up to the individual researcher to decide what assumption best meets the needs of their analysis. The Urban Institute has not yet developed any standard methods for performing such allocations.

Other Issues in Using the CPS to Measure Health Insurance Coverage

For a complete discussion of other issues, see Appendix Three of the "State Level Databook on Health Care Access and Financing", second edition.

Notes:
1) When Census began separating Indian Health Coverage from Medicaid coverage, it decided to classify persons covered only by Indian Health as uninsured in it's publications. Whether we consider such persons as covered or not depends upon the purpose of our analysis.
2) Despite its name, it is unclear from the CPS questionnaire whether the variable HealthCoveredOtherNonGov indicates government or non-government coverage. Currently, in our analysis we consider it to indicate government coverage.
3) Strictly speaking, we do not know whether a person covered from outside the household is covered by a group or non-group plan. However, in our analysis we consider such a person to be covered by a group plan.
4) See Katherine Swartz, 1986, "Interpreting the Estimates from Four National Surveys of the Number of People without Health Insurance", Journal of Economic and Social Measurement 14:233-42.