Welfare Leavers and Medicaid Dynamics: Five States in 1995

By Marilyn Ellwood & Carol Irvin

Mathematica Policy Research, Inc. 50 Church Street, Fourth Floor Cambridge, MA 02138

April 14, 2000

"

Executive Summary

State welfare caseloads have been declining at an unprecedented rate since 1994, partly as a result of state and federal welfare reform efforts and partly because of a strong economy. Medicaid enrollment for children and their parents has been shrinking as well (although less so than welfare), in spite of state efforts to expand their Medicaid eligibility policies. These declines in Medicaid enrollment were not expected, given the many ways family members can remain eligible for Medicaid after welfare (especially children). In addition, there has not been a decline in the number of uninsured.

This study analyzed the 1995 Medicaid enrollment patterns of children and their parents in five states--Alabama, California, Florida, Michigan and New Jersey--to increase our understanding of the enrollment interactions between welfare and Medicaid. Each of these states experienced a decline in their welfare caseloads during 1995. Using Medicaid administrative data, we explored the association of the welfare declines with overall Medicaid enrollment patterns. One part of the analysis focused on persons who left welfare during the year to see how many stayed on Medicaid. We also examined whether welfare leavers who stayed on Medicaid had different Medicaid expenditure patterns from those who left. A second part of the analysis looked at the dynamics of overall Medicaid enrollment during 1995, with a focus on the extent of turnover.

State Medicaid Research Files (SMRF) from the Health Care Financing Administration (HCFA(now known as CMS)) are the primary data source. For each state, these files include monthly eligibility information for all Medicaid enrollees during a calendar year, as well as information about individual expenditure levels.

Findings

Study findings add to the body of evidence that declines in the welfare caseload which began in 1995 are likely to have a noticeable effect on state Medicaid programs, in terms of overall enrollment, caseload mix, and per capita expenditure levels. The findings also point to problems of continuity in Medicaid enrollment, which may be contributing to the lack of insurance among low- income children and their parents. Medicaid enrollment patterns that are cause for concern include:

· Declines in state welfare caseloads are associated with declines in Medicaid enrollment, particularly for parents. Every state showed a decrease in Medicaid enrollment during 1995, among family members qualifying for coverage through welfare. Some of the states were able to increase enrollment through the nonwelfare eligibility groups by making greater use of the child poverty-related, transitional assistance, and medically needy groups. Nevertheless, all the states had a net decline in the Medicaid enrollment of children and their parents. The Medicaid declines ranged from 1 to 4 percent, compared to a 2 to 12 percent decline in welfare caseloads. In every state, declines were greater for adults than children.

· Many welfare leavers are not staying on Medicaid and are at risk of becoming uninsured. At least half the children and adults leaving welfare in every study state also left Medicaid. While we expected that a sizable proportion of parents leaving welfare might not continue on Medicaid, this result was not expected for children, given the availability of the poverty-related expansion coverage in all the states. This pattern is troubling since the work of other researchers suggests that while not all persons leaving welfare and medicaid lose their insurance coverage, a significant number of welfare leavers become uninsured.

· Medicaid programs appear to disproportionately lose low-cost welfare leavers while continuing to cover those with higher costs. Study results show that children and adults who left welfare and Medicaid generally had lower average Medicaid expenditures compared to those who remained enrolled. Findings also showed that adults and children with any amount of Medicaid expenditures were more likely to maintain Medicaid coverage upon leaving welfare. In addition, at least among children, as monthly expenditures increased, so did the probability of remaining enrolled in Medicaid. These findings suggest that the Medicaid casemix may be changing. Medicaid managed care programs are understandably worried that those who remain on Medicaid may have more health problems, and that capitated payments may not adequately reflect casemix changes.

· States are experiencing considerable turnover in their Medicaid caseloads. Turnover in Medicaid enrollment was especially high for adults, ranging from 26 to 40 percent during 1995 across the study states. The annual turnover rates for children went from 21 to 29 percent. Turnover is a problem to the extent that individuals leaving Medicaid continue to be eligible. Other researchers have found that many children who are uninsured were previously enrolled in Medicaid, and that many still appeared to be eligible. The sheer numbers of individuals moving off of Medicaid in 1995 dwarfed the year end declines in enrollment. Across the study states, monthly enrollment had declined about by 180,000 individuals at year end; however, nearly 1.5 million children and almost a million adults disenrolled from Medicaid during the year.

Since state welfare caseloads declined at much greater rates after 1995, there is concern that Medicaid enrollment will drop even more, and that Medicaid reductions may be contributing to the continuing problem of the uninsured. The recently implemented State Child Health Insurance Program (SCHIP) expansions for children (and the associated outreach) provide some hope that these Medicaid enrollment declines eventually will be reversed, and that fewer low-income children will be uninsured. However, the situation for parents seems less likely to improve less unless they receive greater attention.

State officials report that many families leave welfare without providing the information needed to redetermine their continued eligibility for Medicaid. Some families leaving welfare may not need Medicaid, since they will be able to secure private insurance through employment; but others who become uninsured may mistakenly believe that they no longer qualify for Medicaid. It is also likely that some welfare leavers perceive the effort associated with maintaining Medicaid coverage not worth the benefit; some families have indicated that they know they can reapply for Medicaid if they need to.

In one sense, it is reassuring that welfare leavers who also leave Medicaid may be healthier (since they appear to have used fewer medical services prior to welfare exit). However, if these individuals leave welfare and become uninsured, they may jeopardize their future health status if they lack ready access to care and delay needed services. Lack of insurance may keep children, in particular, from obtaining appropriate preventive services.

Congress and the states continue to be concerned about the high rate of uninsured. The results from this study suggest that the dynamics of Medicaid enrollment may be at least part of the problem. If state Medicaid programs could increase the number of children and parents leaving welfare who continue on Medicaid, uninsurance rates might decline. In addition, greater attention needs to be given to inappropriate turnover among children and adults qualifying for Medicaid through the nonwelfare groups. Although states may continue to make policy changes to expand Medicaid coverage provisions in an effort to reduce the number of uninsured, these changes are not likely to be effective unless steps are also taken to improve individual continuity in Medicaid coverage.

Introduction

State welfare caseloads have been declining at an unprecedented rate since 1994, partly as a result of state and federal welfare reform efforts and partly because of a strong economy. From a peak of 14.2 million recipients in 1994, by 1998 monthly welfare enrollment had dropped more than 40 percent to an average of 8.3 million recipients. Medicaid enrollment for children and their parents has been shrinking as well, in spite of state efforts to expand their Medicaid eligibility policies (Ellwood and Ku 1998). The Medicaid declines are much more modest, due to the offsetting coverage available through the poverty-related groups for children and other eligibility provisions not tied to cash assistance. Recent analyses suggest that the number of children enrolled in Medicaid decreased by 2.7 percent from 1995 to 1997, whereas adult (or parental) enrollment decreased by 10.6 percent (Ku and Bruen 1999). This represents a decline in enrollment of about 1.6 million children and adults. States have substantially increased the number of children and their parents qualifying for Medicaid under the noncash eligibility groups, but this increase has not been sufficient to mitigate the declining number who qualify for Medicaid because they meet cash assistance-related rules.

These declines in Medicaid enrollment were not expected. Under the federal welfare reform legislation, called the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996, Congress tried to minimize any adverse effects of welfare reform on Medicaid, leaving in place the 12 months of extended Medicaid coverage for families leaving welfare for work and prohibiting states from imposing work requirements or time limits for Medicaid. The PRWORA legislation also provided states with considerable flexibility to make their Medicaid eligibility rules less restrictive, and PRWORA did not change the continued phase-in of the poverty-related Medicaid expansions for children. In addition, the State Child Health Insurance Program (SCHIP) legislation of 1997 provided enhanced matching funds to states to further accelerate their coverage of uninsured low-income children. Thus, States were enabled to follow several paths in expanding their Medicaid eligibility policies. There are several means by which families leaving welfare should continue to qualify for Medicaid, but to date, these provisions have not been sufficient to stem a decline in Medicaid caseloads. In view of the fact that the number of uninsured according to Current Population Survey data has risen every year since 1987, this trend is causing concern.

An earlier paper by Ellwood and Lewis (1999) used Medicaid administrative data from California and Florida for 1995 to analyze how declines in welfare enrollment were affecting Medicaid. Generally, they found that at least half of children and one-half to two-thirds of parents leaving welfare also left Medicaid, even though a greater proportion would be expected to remain eligible. Ellwood and Lewis found, too, that one-quarter to one-third of the child and adult Medicaid caseloads turned over during the year, and that 7 to 10 percent of the enrollees had short-term interruptions in their Medicaid enrollment during the year (usually referred to as “churning”). Their research suggests that the problem of continuity in Medicaid enrollment existed before the enactment of PRWORA in 1996 and may be a factor in the Medicaid enrollment decline, primarily because a sizable proportion of welfare leavers do not remain on Medicaid.

This study builds on that research by extending the analysis to additional states and examining the relationship between Medicaid expenditure levels for people leaving welfare and the likelihood of continued Medicaid enrollment. Managed care organizations are understandably concerned that the individuals who leave welfare and Medicaid may be healthier, leaving a disproportionate number of those who are sicker and more expensive on Medicaid. Obviously, this could affect the adequacy of capitation rates that are being used under Medicaid managed care programs.

We focus on the 1995 Medicaid caseloads of children and their parents in five states--Alabama, California, Florida, Michigan and New Jersey. Each of these states experienced a decline in their welfare caseloads during 1995 (Table 1). California’s welfare enrollment dropped less than 2 percent, but the decline in the other states ranged from 8 to 12 percent, according to AFDC data. Specific analyses include:

• What happened with Medicaid enrollment in each state in 1995
• The extent to which persons leaving AFDC during 1995 remained on Medicaid, with separate analyses of children and adults¹
• Whether Medicaid expenditure levels differed for welfare leavers who stayed on Medicaid, compared to those who left both programs
• The level of turnover and churning within the Medicaid caseload in each state

State Medicaid Research Files (SMRF) files from the Health Care Financing Administration (HCFA(now known as CMS)) are the primary data source. For each state, these files include monthly eligibility information for all Medicaid enrollees during a calendar year, as well as information about individual expenditure levels. These five states were chosen for study because by 1995 their welfare caseloads had started to decline, and their SMRF data were readily available.

Previous Research

Concerns about Medicaid enrollment patterns are not new. Although Medicaid enrollment of children and their parents increased by just over 60 percent during the period from 1987 to 1995,

TABLE 1
CHANGE IN NUMBER OF AFDC RECIPIENTS IN STUDY STATES,
1993-1996

State	Alabama	Arkansas	California	Florida	Michigan	New Jerse
	Number of Recipients
Jan. 1993	141,746	73,982	2,415,121	701,842	686,356	349,902
Jan. 1994	135,096	70,563	2,621,383	689,135	672,760	334,780
Jan. 1995	121,837	65,325	2,692,202	657,313	612,224	321,151
Jan. 1996	108,269	59,223	2,648,772	575,553	535,704	293,833
	Percent Change from Previous Year
Jan. 1994	-4.7	-4.6	8.5	-1.8	-2.0	-4.3
Jan. 1995	-9.8	-7.4	2.7	-4.6	-9.0	-4.1
Jan. 1996	-11.1	-9.3	-1.6	-12.4	-12.5	-8.5
Source: U.S. Department of Health and Human Services, Administration for Children and Families. 1995 State Medicaid Research File data.

research during that period indicated that many seemingly eligible individuals were not enrolled in Medicaid, and that individuals were leaving Medicaid without insurance. Generally, lower participation rates were reported for persons eligible under the nonwelfare routes to Medicaid eligibility. For 1993, Dubay and Kenney (1996) estimated a participation rate of 69 percent for uninsured children who would qualify under the child poverty-related expansion provisions, compared to a 90 percent rate for children qualifying under the welfare-related rules. Ellwood and Adams (1990) found that the majority of cases leaving Aid to Families with Dependent Children (AFDC) in California and Georgia were not continuing on Medicaid, despite the availability of transitional coverage for families going from welfare to work. Short and Freedman (1998), using data from the 1990 panel of the Survey of Income and Program Participation (SIPP), found that two- thirds of mothers who left Medicaid became uninsured. Short and Freedman also reported that 10 percent of the Medicaid caseload they were studying turned over in a four-month period.

Studies using more recent data confirm the persistence of Medicaid enrollment problems and raise other concerns as well. Selden et al. (1998), using 1996 data from the Medical Expenditure Panel Survey, found that 4.7 million children in 1996 were uninsured despite being eligible for Medicaid. Using a national survey of welfare leavers from 1995 to 1997, Garrett and Holahan (2000 a and b) found that 44 percent of mothers and 30 percent of children who left welfare less than six months earlier were not enrolled in Medicaid. They also found that rates of uninsurance for families leaving welfare increased with the number of months since leaving welfare. Important to this study, a person’s health status, they found, was important in determining whether Medicaid coverage was maintained after leaving welfare. Women who had health conditions that limited their ability to work were more likely to stay on Medicaid after leaving welfare, compared to women without limiting health conditions. Results were similar, though weaker, for children. These findings are important because they suggest that welfare leavers remaining on Medicaid are likely to have greater health care needs and incur greater expenses than those who leave both welfare and Medicaid.

Theories abound as to why eligible families do not participate in Medicaid, including families who may have had some period of enrollment in the past. Focus groups and surveys have identified the following barriers to participation (Smith et al. 1998; Perry et al. 1998; and Shuptrine et al. 1998):

• The stigma associated with welfare receipt extends to Medicaid as well, and keeps many families from applying for coverage or continuing on Medicaid once they are off welfare.
• Many low-income people think Medicaid is for families on welfare, not working families.
• Some people confuse the new rules associated with welfare reform with Medicaid rules, leading them to believe mistakenly that Medicaid is now time-limited, like welfare, or that Medicaid coverage (without welfare) counts as part of the new welfare lifetime limit, or that the welfare work requirements extend to Medicaid.
• Immigrants, in particular, are worried that participation by any family member in Medicaid (even children who are citizens) may cause parents to be considered public charges and thus disqualify them from eventual citizenship.
• Families who have been on welfare and Medicaid in the past say they dropped out because the eligibility process is burdensome and demeaning, or they were frustrated by the complexity of the rules.
• Families say they are healthy, and they believe they can get Medicaid if they need it.

Thus, even when states make their Medicaid eligibility policies more generous, other barriers to participation may prevent the expansions from maintaining or increasing enrollment.

Eligibility Policies of Study States in 1995

By 1995, interstate differences in Medicaid eligibility policy had been considerably reduced as a result of the federally mandated poverty-related expansions for children. That same year, all states were required to extend Medicaid coverage to children under age six and pregnant women with family income below 133 percent of the federal poverty level (FPL) and to children born after September 30, 1983, with family income below 100 percent FPL. The study states exceeded these federal requirements only with a few groups in 1995 (Table 2). Four of the states (California, Florida, Michigan, and New Jersey) used income thresholds ranging from 185 to 200 percent FPL for infants and pregnant women, and Michigan covered children ages 1 through 15 years to 150 percent FPL.

Nevertheless, there continued to be substantial variation among the study states with their AFDC and medically needy income levels, which in particular affected Medicaid eligibility for adults. California’s AFDC benefit levels were at 58 percent of FPL, among the highest in the United States. In contrast, the AFDC levels of Alabama and Florida at 15 to 29 percent of FPL were considerably less generous and well below the median state. Among the study states, Alabama did not have medically needy coverage in 1995, and New Jersey’s medically needy program was somewhat restrictive, since it covered children and pregnant women but no other parents. The medically needy income thresholds ranged from 29 to 78 percent FPL, with those of all the states but California below the median state.

All the states were required to extend up to 12 months of transitional Medicaid coverage to children and their parents if families were leaving welfare due to earnings. All the study states also used other more obscure, noncash eligibility groups to establish eligibility for Medicaid-only benefits for families not poor enough to qualify for welfare.²

TABLE 2
MEDICAID INCOME THRESHOLDS FOR STUDY STATES, 1995
(Percentages of the Federal Poverty Level)

	Alabama	Arkansas	California	Florida	Michigan	New Jersey	U.S.a
AFDC	15	19	58	29	45	41	36
Medicaid needy	--	25	78	29	52	52	56
Poverty-related
Infants/pregnant women	133	133	200	185	185	185	133b
Children ages 1-5	133	13	133	133	150	133	133b
Children ages 6-12c	100	100	100	100	150	100	100b
Children ages 13-18	None	None	None	None	150d	None	None
Source:  MCH Update.  State Medicaid Coverage of Pregnant Women and Children: Summer 1995.  Washington, D.C.: National Governors' Association. aMedian state. bMinimum federal requirements. cChildren born after September 30, 1983, which, for most of 1995, covered children through age 12. dChildren under age 15 in Michigan were covered to 150 percent FPL.

It is difficult to predict exactly which eligibility groups will be important to a state in extending Medicaid coverage. Children and other family members often can qualify under more than one Medicaid eligibility provision. Which provision is used may depend on numerous factors, including whether the family reported earnings, the level of family income, whether any family member had unusually high medical expenses, and the ages of the children in the family. Other factors include the extent of automation of eligibility determination in each state and the structure and internal logic of the automated system. For example, the systems in many states employ a hierarchical approach to eligibility determination, testing families for cash assistance and Medicaid first (using the lowest income thresholds) and poverty-related coverage last (using the highest income thresholds), with the other eligibility groups (such as transitional assistance and medically needy coverage) somewhere in between.

Results

Did Medicaid Enrollment Decline?

Consistent with the national pattern, all five study states experienced a decline in Medicaid enrollment for children and adults during 1995, although the extent of the decline varied (Table 3). New Jersey’s Medicaid enrollment in December was only 0.8 percent below its January level, compared to a 4.2 difference in Florida by year end. The other three states had declines of 2.1 to 3.8 percent.

TABLE 3
DISTRIBUTION BY ELIGIBILITY GROUP OF MEDICAID ENROLLMENT
FOR STUDY STATES, JANUARY AND DECEMBER 1995

	States
	Alabama	Arkansas	California	Florida	Michigan	New Jersey
Enrollment in Jan.1995
AFDC	126,730	69,617	2,690,191	702,315	612,443	326,007
Poverty-related	142,368	52,908	80,685	258,731	100,091	67,562
Medically Needy		10,747	697,860	18,668	74,545	3,086
Other	30,057	14,557	643,006	167,244	64,257	88,404
Total	299,155	147,829	4,110,742	1,146,958	851,336	485,059
Enrollment in Dec.1995
AFDC	113,595	61,413	2,651,487	649,143	537,089	303,281
Poverty-related	146,907	53,254	99,109	261,568	123,502	71,076
Medically Needy		9,549	680,356	17,973	76,619	2,594
Other	28,936	13,759	592,085	170,399	82,117	104,174
Total	289,438	137,975	4,023,037	1,099,083	819,327	481,125
Percent Difference
AFDC	-10.4	-11.8	-1.4	-7.6	-12.3	-7.0
Poverty-related	3.2	0.7	22.8	1.1	23.4	5.2
Medically Needy		-11.1	-2.5	-3.7	2.8	-15.9
Other	-3.7	-5.5	-7.9	1.9	27.8	17.8
Total	-3.2	-6.7	-2.1	-4.2	-3.8	-0.8
Children	-2.0	-5.2	-0.5	-2.3	-3.0	-0.7
Adults	-11.7	-14.8	-5.6	-9.6	-8.4	-2.2
Source:  1995 State Medicaid Research File data.

The level of Medicaid decline attributable to reduced AFDC enrollment was fairly consistent with the level of decline reported by each state’s welfare program (shown in Table 1).(3) Helping offset the welfare declines in every state were increases in the poverty-related groups, although the extent of the increase varied. The increase in poverty-related coverage was only 3.2 percent in Alabama, but this lower rate occurred in part because the poverty-related group already accounted for nearly half the state’s Medicaid enrollees at the beginning of the year. At the other extreme, poverty-related coverage increased by almost 23 percent during 1995 in California, but it still accounted for only 2.4 percent of overall enrollment at year’s end. Michigan also had a high growth rate (23.4 percent) for poverty-related coverage during the year. By year’s end, the poverty-related groups accounted for 15.1 percent of Medicaid enrollment in Michigan. Alabama and Florida, which had the lowest AFDC benefit levels and no or marginal medically needy coverage among the study states, had the lowest rates of increase for poverty-related coverage, but they were making greater use of the poverty-related coverage at the start of 1995 than the other three states.

There were no clear changes in enrollment patterns across the states for persons covered under either the medically needy group or the residual “other” eligibility group. Michigan was the only state with growth in medically needy enrollment during 1995. Michigan and New Jersey reported sizable increases in enrollment under the other coverage group (27.8 percent and 17.8 percent, respectively). This residual coverage group includes families qualifying for transitional assistance. California’s 7.9 percent decrease in enrollment under the other coverage group occurred because of a decline during the year in the number of families affected under a class-action court decision.⁴

In all five states, the drop in adult Medicaid enrollment during 1995 was greater than that reported for children. This pattern was expected, since adults in all the states had fewer options for noncash coverage than children.

Did Welfare Leavers Stay on Medicaid?

A key part of our analysis was to see what happened to the Medicaid status of individuals leaving AFDC. We focused on those who left from February through July 1995, so that we could follow the Medicaid status for six months after AFDC exit. Six months seemed to us a sufficient time to assess whether Medicaid coverage continued. We counted persons as having left AFDC only if they did not receive benefits for two consecutive months. This standard reduced the likelihood of counting persons whose AFDC coverage was interrupted temporarily because of failure to comply with reporting or other administrative requirements.

What proportion of adults who left AFDC stayed on Medicaid? Across the five states, from 49.3 percent (California) to 66.6 percent (Florida) of adults were no longer enrolled in Medicaid three months after leaving AFDC (Table 4). This overall pattern changed very little at the six-month interval. At six months, three of the states had a slight increase in the proportion of adults who were covered by Medicaid, but only because more adults had returned to the AFDC rolls (and were thus automatically reenrolled in Medicaid). Transitional benefits were used by a relatively small proportion of adults in the three states in which data on transitional coverage were available, ranging from 8.4 percent of adults in California to 19.1 percent in New Jersey. By the sixth month after AFDC exit, between 12 percent (New Jersey) and 18.8 percent (California) of adults across the study states had returned to the AFDC program.

TABLE 4
PERCENT DISTRIBUTION OF ADULTS LEAVING AFDC FROM FEBRUARY THROUGH JUNE 1995,
BY THEIR MEDICAID STATUS THREE AND SIX MONTHS AFTER LEAVING AFDC

	Alabama	Arkansas	California	Florida	Michigan	New Jersey
	Three Months After Leaving AFDC
Not Enrolled in Medicaid	59.0	71.8	49.3	66.6	50.0	62.8
Medicaid Enrollment Group
Transitional Coverage	NAa	17.3	8.4	14.2	NAa	19.1
Medically Needy	--	1.8	7.6	2.2	7.2	<0.0
Poverty-related	0.4	1.5	0.3	1.8	1.5	0.2
SSI	3.9	1.8	2.3	2.7	6.2	1.4
Other Non-cash	31.0	2.6	22.3[1]	6.0	30.1	12.2
AFDC Cash	5.7	3.2	9.9	6.5	5.0	4.2
Subtotal	41.0	28.2	50.7	33.4	50.0	37.2
Total	100.0	100.0	100.0	100.0	100.0	100.0
	Six Months After Leaving AFDC
Not Enrolled in Medicaid	60.5	65.2	56.2	64.8	49.1	61.2
Medicaid Enrollment Group
Transitional Coverage	NAa	13.1	6.1	8.6	NAa	18.9
Medically Needy	--	2.0	8.6	1.5	5.5	<0.0
Poverty-related	2.3	1.6	0.3	1.5	1.2	0.3
SSI	3.8	2.0	2.4	2.8	6.7	1.4
Other Non-cash	16.8	1.8	7.4b[2]	3.7	24.1	6.2
AFDC Cash	16.6	14.3	18.8	17.0	13.3	12.0
Subtotal	39.5	34.8	43.6	35.1	50.8	38.8
Total	100.0	100.0	100.0	100.0	100.0	100.0
	(n = 16,432)	(n = 9,137)	(n = 170,726)	(n = 98,717)	(n = 28,692)	(n = 24,544)
Source: SMRF Data NA - Not available aIn Alabama and Michigan, persons qualifying for Medicaid under the transitional coverage provisions could not be separately identified.  For these two states, persons with transitional coverage are included under the Other Non-Cash group. bIncludes court-ordered Edwards v. Kizer group for California, which allows families leaving welfare to remain on Medicaid until their eligibility under other coverage provisions can be redetermined. * Includes court-ordered Edwards v. Kizer group for California which allows families leaving welfare to remain on  Medicaid until their eligibility under other coverage provisions can be redetermined. [1] David Ellwood: [2] David Ellwood:

As we expected, children leaving AFDC were more likely than their parents to remain on Medicaid (Table 5). The margin of difference, however, was not large, except in Florida. Three months after leaving AFDC, the proportion of children not enrolled in Medicaid ranged from 46.7 percent (California) to 57.9 percent (Florida) across the five states. Similar to adults, the situation improved somewhat by the sixth month, but only because more children had returned to the AFDC program. The eligibility groups used to continue coverage varied across the states. California and New Jersey were less likely to use the poverty-related provisions to continue child coverage after AFDC than the other study states. However, it is difficult to compare the states on the basis of which eligibility groups they used to continue Medicaid eligibility. For example, California’s medically needy income threshold was considerably higher than the AFDC threshold, so that more children could qualify under the medically needy provisions, before they were tested for coverage under the poverty-related group. If California's medically needy income threshold had been closer to the AFDC threshold, more children would likely have qualified under the poverty-related provisions.

Thus, in all the study states, close to half or more of children and their parents who left welfare in 1995 also left Medicaid. We cannot be certain what happened to the insurance status of each of these individuals; however, previous research suggests that many became uninsured. In addition, it seems likely that the vast majority of children could have retained their Medicaid eligibility, given the poverty-related provisions available in all the states.

TABLE 5
PERCENT DISTRIBUTION OF CHILDREN LEAVING AFDC FROM FEBRUARY THROUGH JUNE 1995,
BY THEIR MEDICAID STATUS THREE AND SIX MONTHS AFTER LEAVING AFDC

	Alabama	Arkansas	California	Florida	Michigan	New Jersey
	Three Months After Leaving AFDC
Not Enrolled in Medicaid	55.7	64.3	46.7	49.8	46.9	57.9
Medicaid Enrollment Group
Transitional Coverage	NAa	19.4	7.4	15.0	NAa	16.9
Medically Needy	--	2.6	8.4	0.8	10.6	<0.0
Poverty-related	0.7	5.7	0.9	18.0	9.7	1.5
SSI	3.5	1.1	1.4	1.9	<0.0	1.3
Other Non-cash	35.3	3.5	23.4b	7.7	27.5	18.4
AFDC Cash	4.8	3.5	12.0	7.0	5.3	3.9
Subtotal	44.3	35.7	53.3	50.2	53.1	42.1
Total	100.0	100.0	100.0	100.0	100.0	100.0
	Six Months After Leaving AFDC
Not Enrolled in Medicaid	55.1	56.2	51.0	48.2	44.1	56.9
Medicaid Enrollment Group
Transitional Coverage	NAa	14.2	5.3	9.0	NA	16.4
Medically Needy	--	3.0	9.4	0.6	8.8	<0.0
Poverty-related	7.1	7.1	1.4	15.6	10.1	2.2
SSI	3.7	1.2	1.4	2.1	0.1	1.3
Other Non-cash	20.5	2.5	10.2b	5.6	22.2	11.4
AFDC Cash	13.7	15.9	21.3	19.0	14.8	11.8
Subtotal	45.0	43.9	49.0	51.9	56.0	43.1
Total	100.0	100.0	100.0	100.0	100.0	100.0
	(n = 30,676)	(n = 20,886)	(n = 268,897)	(n = 170,161)	(n = 85,891)	(n = 42,655)
Source: 1995 State Medicaid Research File data. NA - Not available aIn Alabama and Michigan, persons qualifying for Medicaid under the transitional coverage provisions could not be separately identified.  For these two states, persons with transitional coverage are included under the Other Non-Cash group. bIncludes court-ordered Edwards v. Kizer group for California, which allows families leaving welfare to remain on Medicaid until their eligibility under other coverage provisions can be redetermined. * Includes court-ordered Edwards v. Kizer group for California which allows families leaving welfare to remain on  Medicaid until their eligibility under other coverage provisions can be redetermined.

Were the Welfare Leavers Staying on Medicaid Different From Those Who Left?

Policymakers and Medicaid officials are understandably interested in whether welfare leavers who also leave Medicaid are different from those who remain enrolled in Medicaid. In particular, they want to know whether welfare leavers who continue on Medicaid are likely to be more costly than those who leave Medicaid would have been. Differences in expected expenditure levels could cause State Medicaid programs to experience adverse selection among those leaving welfare. Welfare leavers with few medical-care needs at the time of welfare exit may be less inclined to maintain Medicaid coverage, particularly if they believe that they will be able to reestablish Medicaid coverage if they need it. Conversely, welfare leavers with medical needs--especially those with considerable needs--can be expected to maintain Medicaid coverage in order to maintain access to care. As a result, State Medicaid programs may be continuing coverage for welfare leavers with greater anticipated medical expenditures, while losing healthier welfare leavers who are likely to cost less.

SMRF data enabled us to explore whether there were differences in expenditure patterns among welfare leavers who left Medicaid, compared to those who remained enrolled. We first looked at unadjusted expenditure differences and then differences that adjusted for age, sex, and race/ethnicity in a regression framework. However, the SMRF expenditure data had important limitations. First, we could only use expenditure data for welfare leavers not participating in managed care programs. Because services provided through managed care are not available in the SMRF data, the expenditure information for these individuals is not equivalent to those in fee-for-service. We were able to identify those individuals in managed care in all the study states but California; therefore, California is excluded from the expenditure analysis.⁵ Second, the person summary files in the SMRF data only included annual expenditure levels. As a result, we could not separate Medicaid expenditures for individuals into the time periods before and after welfare exit. To adjust for varying lengths of enrollment, we computed an average monthly expenditure level.

The sample used in the previous analysis was further restricted for this analysis of expenditures. First, we limited our sample for the expenditure analysis to persons who left welfare in July through October. Our earlier analysis also included persons who left welfare in February through June. This restriction allowed us to minimize the time period over which expenditures could be incurred after welfare exit. Nevertheless, some bias to our results is possible if the level of Medicaid expenditures after welfare exit (for those who remained on Medicaid) was markedly different from the expenditure level prior to exit. Second, the sample for the expenditure analysis was further restricted to welfare leavers who were enrolled continuously until the point of welfare exit in order to minimize the problem of enrollment periods that were too short for individuals to obtain medical care. As a result, all individuals in our sample had at least six months of enrollment during 1995.

As we expected, study results showed that most individuals who left welfare, but did not maintain Medicaid coverage, had lower average monthly expenditures compared to those who remained on Medicaid. Unadjusted data are presented in Tables 6 and 7. These data show that adults and children who left welfare but did not maintain Medicaid coverage, were less likely to have had at least some Medicaid expenditures when they were on welfare; they were less likely to have had an inpatient admission; their median monthly Medicaid expenditures were lower; and most had average monthly expenditures that were lower than the expenditures of welfare leavers who remained on Medicaid.

TABLE 6
MEDICAID EXPENDITURES FOR ADULTS LEAVING AFDC
FROM JULY THROUGH OCTOBER 1995
(Adults in HMOs Excluded)

	Alabama	Florida	Michigan	New Jersey
Month 3 after AFDC:  Not enrolled in Medicaid
Percent with Zero Expenditures	35.0***	39.1***	22.5***	30.6***
Percent with an Inpatient Admission	7.4***	8.2***	10.0***	9.8***
Median Monthly Expenditures a	$19	$19	$36	$36
Average Monthly Expenditures a	$123***	$145***	$164	$280***
Month 3 after AFDC:  Enrolled in Medicaid
Percent with Zero Expenditures	8.8	12.2	5.5	6.0
Percent with an Inpatient Admission	15.4	16.8	14.6	13.6
Median Monthly Expenditures a	$46	$69	$52	$64
Average Monthly Expenditures a	$139	$203	$158	$235
SOURCE:  1995 State Medicaid Research File data. aBased on number of months enrolled. *Significantly different from adults maintaining Medicaid coverage after leaving welfare at the .10 level, two-tailed test. **Significantly different from adults maintaining Medicaid coverage after leaving welfare at the .05 level, two-tailed test. ***Significantly different from adults maintaining Medicaid coverage after leaving welfare at the .01 level, two-tailed test.

TABLE 7
MEDICAID EXPENDITURES FOR CHILDREN LEAVING AFDC
FROM JULY THROUGH OCTOBER 1995
(Children in HMOs Excluded)

	Alabama	Florida	Michigan	New Jersey
Month 3 after AFDC:  Not enrolled in Medicaid
Percent with Zero Expenditures	43.9***	46.0***	30.3***	38.6***
Percent with an Inpatient Admission	2.3***	1.9***	2.8***	2.5***
Median Monthly Expenditures a	$7	$6	$14	$11
Average Monthly Expenditures a	$49***	$61***	$81	$84**
Month 3 after AFDC:  Enrolled in Medicaid
Percent with Zero Expenditures	10.5	12.0	6.3	8.2
Percent with an Inpatient Admission	6.7	4.0	4.8	5.6
Median Monthly Expenditures a	$26	$32	$24	$29
Average Monthly Expenditures a	$95	$92	$66	$99
SOURCE:  1995 State Medicaid Research File data. aBased on number of months enrolled. *Significantly different from children maintaining Medicaid coverage after leaving welfare at the .10 level, two-tailed test. **Significantly different from children maintaining Medicaid coverage after leaving welfare at the .05 level, two-tailed test. ***Significantly different from children maintaining Medicaid coverage after leaving welfare at the .01 level,  two-tailed test.

Among the adults who left both welfare and Medicaid, the proportion incurring no Medicaid expenditures while enrolled ranged from 22.5 percent (Michigan) to 35.0 percent (Alabama). In comparison, only 5.5 percent (Michigan) to 12.2 percent (Florida) of the adult welfare leavers who remained on Medicaid had zero Medicaid expenditures while enrolled. Inpatient admissions ranged from 7.4 percent (Alabama) to 10.0 percent (Michigan) among those leaving Medicaid, and from 13.6 percent (New Jersey) to 16.8 percent (Florida) among adults maintaining coverage. Median monthly Medicaid expenditures were less among those leaving Medicaid and ranged from $19 (Alabama and Florida) to $36 (Michigan and New Jersey). Adults maintaining Medicaid coverage had median monthly expenditures that ranged from $46 (Alabama) to $69 (Florida). Average monthly Medicaid expenditures for adults leaving welfare and Medicaid were significantly less in two of the four states and ranged from $123 (Alabama) to $280 (New Jersey), compared to a range of $139 (Alabama) to $235 (New Jersey) among adults who remained on Medicaid after leaving welfare.⁶

Expenditure patterns among children were similar to those of adults: children who left Medicaid were less likely to have expenditures while enrolled and less likely to have an inpatient admission; they had lower median and average monthly Medicaid expenditures. From 30.3 percent (Michigan) to 46.0 percent (Florida) of children leaving welfare and Medicaid did not incur any expenditures while enrolled. This compares to a range of 6.3 percent (Michigan) to 12.0 percent (Florida) for those who maintained their Medicaid coverage. Across the study states, between two and three percent of children leaving Medicaid had inpatient hospital admissions prior to welfare exit, compared to a rate of four to seven percent among children who remained on Medicaid. Average monthly Medicaid expenditures for children leaving welfare and Medicaid were significantly less in three of the four states (the exception was Michigan), ranging from $49 (Alabama) to $84 (New Jersey), compared to a range of $66 (Michigan) to $99 (New Jersey) among children who remained on welfare after Medicaid.

We also analyzed the relationship between monthly Medicaid expenditures and whether an individual maintained Medicaid after leaving welfare in a regression framework. Logistic regression was used to determine whether greater monthly Medicaid expenditures were associated with an increased probability that a welfare leaver would maintain Medicaid coverage. Tables 8 and 9 present the results of the regression analysis and Appendix Tables 1 and 2 present the full regression specification. The analysis is based on comparing each quintile of expenditures to having no expenditures. The use of quintiles rather than a continuous measure of average monthly expenditures allowed us to look at sections of the distribution of expenditures and compare and contrast those with few expenditures and in the first quintile to those with relatively large expenditures and in the fifth quintile.⁷

Similar to the unadjusted data, across all states adults and children with any positive amount of monthly Medicaid expenditures while enrolled were significantly more likely to maintain Medicaid coverage when leaving welfare. For example, in Alabama, adults in the first quintile of expenditures were two times more likely to maintain coverage compared to adults with no expenditures. If an adult had expenditures in the fifth quintile, the probability of maintaining coverage was 2.6 times.

TABLE 8
ESTIMATED ODDS RATIOS OF MEDICAID EXPENDITURES AND THE PROBABILITY
OF ADULTS MAINTAINING MEDICAID COVERAGE WHEN LEAVING AFDC,
COMPARED TO THE PROBABILITY OF LEAVING MEDICAID
(Adults in HMOs Excluded)

	Alabama	Florida	Michigan	New Jersey
Enrolled in Medicaid Three Months After AFDC a
Average Monthly Expenditures b
First quintile	2.02***	2.20***	1.95***	3.47***
Second quintile	2.21***	2.92***	1.99***	3.50***
Third quintile	2.57***	3.64***	2.02***	3.36***
Fourth quintile	2.81***	4.05***	2.19***	3.21***
Fifth quintile	2.63***	3.45**	1.56***	2.40***
SOURCE:  1995 State Medicaid Research File data, adults leaving welfare from July through October, 1995. aCompared to leaving AFDC and Medicaid. bCompared to no expenditures. *Significantly different from zero at the .10 level, two-tailed test. **Significantly different from zero at the .05 level, two-tailed test. ***Significantly different from zero at the .01 level, two-tailed test.

TABLE 9
ESTIMATED ODDS RATIOS OF MEDICAID EXPENDITURES AND THE PROBABILITY
  OF MAINTAINING MEDICAID COVERAGE WHEN LEAVING AFDC,
COMPARED TO THE PROBABILITY OF LEAVING MEDICAID
(Children in HMOs Excluded)

	Alabama	Florida	Michigan	New Jersey
Enrolled in Medicaid Three Months after AFDC a
Average Monthly Expenditures b
First quintile	4.26***	3.21***	3.03***	3.76***
Second quintile	5.15***	3.83***	3.44***	3.67***
Third quintile	4.94***	4.49***	3.68***	4.14***
Fourth quintile	6.84***	4.91***	3.67***	4.37***
Fifth quintile	9.16***	5.53***	4.22***	4.67***
SOURCE:  1995 State Medicaid Research File data, children leaving from July through October 1995. aCompared to leaving AFDC and Medicaid. bCompared to no expenditures. *Significantly different from zero at the .10 level, two-tailed test. **Significantly different from zero at the .05 level, two-tailed test. ***Significantly different from zero at the .01 level, two-tailed test.

A similar pattern is seen among children--when they had at least some Medicaid expenditures while enrolled, they were significantly more likely to maintain Medicaid coverage and the magnitude of a child’s monthly expenditures was more likely to be related to the probability of maintaining Medicaid coverage. For example, in Alabama, children with expenditures in the first quintile were more than four times as likely to maintain Medicaid coverage compared to children with no expenditures. When children in Alabama had expenditures in the fifth quintile, they were nine times more likely to maintain Medicaid when leaving welfare. Clearly, in Alabama whereas having any expenditures significantly increased the probability that a child leaving welfare would maintain Medicaid coverage, children with greater levels of expenditures were much more likely to maintain coverage, compared to those with fewer expenditures.

How Much Turnover and Churning Occurs in Medicaid Caseloads?

A final study objective was to look at the extent to which turnover and churning in each state’s Medicaid caseload might be contributing to enrollment declines. As background for understanding these additional measures, it is useful to review the different ways Medicaid enrollment is counted. Most analyses of trends in Medicaid enrollment rely on year-to-year counts of the number of persons ever enrolled during the year, which states report to HCFA(now known as CMS) on Form 2082. These aggregate annual counts of the persons ever enrolled in Medicaid are a good measure of what is happening to overall enrollment, but they provide only limited information. Important to this discussion, annual counts do not reveal the extent to which persons remain enrolled from one year to the next. It is theoretically possible that individuals enrolled one year differ significantly from those enrolled the next year. There may be a tendency with Form 2082 data to think that most of the people enrolled from year to year are the same, but this may be erroneous.

Annual counts of the ever-enrolled, like those used in HCFA(now known as CMS)'s Form 2082, are derived by adding the number of persons enrolled at the start of the year to the number of new entries throughout the year. The level of exits or turnover during a given year do not affect this number. In addition, a person enrolled 1 month is counted the same as a person enrolled all 12 months. Those exiting the program during a year have an effect on the next year’s enrollment numbers, but only to the extent that fewer people are enrolled at the start of the next year.

As reported earlier, the SMRF data used for this study allowed us to look at monthly Medicaid enrollment counts for each state, similar to those used to track enrollment in state AFDC/TANF programs. Monthly counts are compiled as a net of persons entering, exiting, and continuing enrollment each month. We think they are a better barometer of enrollment, since they measure a shorter time period and immediately take exits into account. For analysis, it is helpful to have both sets of numbers.

For our analysis, we chose to measure turnover by comparing average monthly enrollment for each state during 1995 to the number of persons ever enrolled in Medicaid at any point during the year. We calculate a turnover rate for each state, based on the difference between the average monthly enrollment and the annual number ever enrolled. The closer a state’s average monthly enrollment is to the annual ever enrolled number, the lower the turnover rate. Among the study states, Florida experienced the greatest turnover in its Medicaid programs in 1995 (Table 10). In Florida, the average monthly enrollment was 68.3 percent of the annual ever-enrolled count. Thus, almost one-third of Florida’s Medicaid caseload turned over during the year. The other four states reported average monthly enrollment ranging from 73 to 78 percent of the annual count, for a turnover rate closer to 25 percent. As expected, in all the states, turnover was greater for adults than children.

TABLE 10
MONTHLY ENROLLMENT VERSUS PERSONS EVER ENROLLED
IN MEDICAID DURING 1995

	Alabama	Arkansas	California	Florida	Michigan	New Jersey
Average Monthly
AFDC cash	119,412	65,093	2,676,904	670,660	575,595	303,281
Poverty-related	145,758	53,565	87,986	259,496	65,411	69,664
Medically needy	NA	10,476	695,562	19,053	76,900	2,894
Other	29,840	14,149	616,437	167,838	74,779	97,585
All persons	295,010	143,283	4,076,889	1,117,048	842,684	485,409
Ever Enrolled
AFDC cash	158,335	94,971	3,106,304	910,522	692,297	384,861
Poverty-related	205,398	87,801	163,561	400,298	114,658	112,083
Medically needy	NA	18,846	1,144,841	57,667	113,750	4,959
Other	39,991	22,067	887,585	266,375	95,500	136,060
All persons	403,724	223,685	5,302,291	1,634,862	1,084,205	637,963
Average Monthly as a Percent
of Ever Enrolled
AFDC cash	75.4	68.5	86.2	73.7	83.1	78.8
Poverty-related	71.0	61.0	53.8	64.8	57.0	62.2
Medically needy	NA	55.6	60.8	33.0	67.6	58.4
Other	74.6	64.1	69.5	63.0	78.3	71.7
All persons	73.1	64.1	76.9	68.3	77.7	76.1
Children	77.3	65.5	78.5	71.4	79.3	78.6
Adults	61.1	51.1	73.5	60.6	74.4	70.6
Annual Turnover Rate
AFDC cash	24.6	31.5	13.8	26.3	16.9	21.2
Poverty-related	29.0	39.0	46.2	35.2	43.0	37.8
Medically needy	NA	44.4	39.2	67.0	32.4	41.6
Other	25.4	35.9	30.5	37.0	21.7	28.3
All persons	26.9	35.9	23.1	31.7	22.3	23.9
Children	22.7	34.5	21.5	28.6	20.7	21.4
Adults	38.9	48.9	26.5	39.4	25.6	29.4
Source:  State Medicaid Research Files

TABLE 4 - California
Enrollment Pattern by Enrollment Group

Principal Enrollment Group	Single Coverage Gap	Other Enrolled in December	Other Not Enrolled in December
AFDC Cash	107,640	12,621	7,443	127,704	afdc
Transitional AFDC	3,509	232	899	4,640
EvM	10,975	829	2,906	14,710
Poverty-Related Eligibles	11,184	5,844	6,466	23,494	pov
Medically Needy	82,837	34,940	37,986	155,763	mn
Other	25,728	9,191	9,428	44,347
Disabled	1,611	384	213	2,208
All Persons	243,484	64,041	65,341	372,866
				65,905	other
				221,668

TABLE 4 - Florida
Enrollment Pattern by Enrollment Group -Number of Enrollees

Principal Enrollment Group	Single Coverage Gap	Other Enrolled in December	Other Not Enrolled in December
AFDC Cash	59,692	14,319	13,173	87,184	afdc
Transitional AFDC	5,279	707	2,063	8,049
Poverty-Related Eligibles	31,158	7,048	6,494	44,700	pov
Medically Needy	1,681	2,383	5,551	9,615	mn
Other	8,265	3,251	4,267	15,783
Disabled	1,098	298	124	1,520
				25,352	other
All Persons	107,173	28,006	31,672

Turnover is appropriate when persons exiting the program are no longer eligible. Usually, turnover is thought of as inappropriate when eligible persons leave the program. Because state Medicaid data do not include information on why people disenroll from Medicaid, we have no way of knowing the extent of turnover that may be inappropriate. However, other researchers have shown that, although some children and adults leaving welfare become insured, a sizable proportion become uninsured.

Related to the turnover patterns, more than half the children in Alabama, California, Michigan, and New Jersey were enrolled throughout 1995 (Table 11). California, Michigan, and New Jersey also had close to half their adult Medicaid population enrolled throughout the year. Average lengths of enrollment for 1995 ranged from an average of 8.6 months for children in Florida, to 9.5 months for children in Michigan. The average length of enrollment during 1995 for adults ranged from 7.3 months in Alabama and Florida, to 8.9 months in Michigan. At the same time, it is important to remember that data on length of enrollment are restricted to 1995 and do not reflect the true average duration of Medicaid enrollment.⁸

Further detail is provided in Table 12, which compares average monthly entry rates (the proportion of persons currently enrolled in Medicaid who were not enrolled the previous month) and exit rates (the proportion of persons not currently enrolled in Medicaid who were enrolled the previous month) for the general population and for the poverty-related child group in 1995. In every state, the average monthly exit rate exceeded the entry rate, as we expected, given the overall decline in enrollment during the year. Also, as expected, given its higher turnover rates, Florida showed the highest rates of entry and exit for the general Medicaid population.

TABLE 11
PERCENT ENROLLEES WITH FULL YEAR ENROLLMENT AND
AVERAGE LENGTH OF ENROLLMENT DURING 1995

	Alabama	Arkansas	California	Florida	Michigan	New Jersey
Percent Enrolled Throughout 1995
Children	52.4	36.2	56.3	43.3	56.7	56.9
Adults	25.2	16.4	48.4	27.8	48.1	46.6
All Persons	45.4	31.2	53.8	39.1	53.4	53.7
Average Length of Enrollment (In Months)
Children	9.3	8.2	9.4	8.6	9.5	9.4
Adults	7.3	6.3	8.8	7.3	8.9	8.5
All Persons	8.8	7.7	9.2	8.2	9.3	9.1
Source:  1995 State Medicaid Research File data.

TABLE 12
ENTRIES TO AND EXITS FROM MEDICAID DURING 1995
AS A PERCENT OF MONTHLY MEDICAID ENROLLMENT

	Alabama	Arkansas	California	Florida	Michigan	New Jersey
Average monthly entries to and exits from Medicaid
Entries	4.0	6.1	3.5	5.4	3.4	3.3
Exits	-4.3	-6.7	-3.7	-5.8	-3.7	-3.4
Net difference	-0.3	-0.6	-0.2	-0.4	-0.3	-0.1
Average monthly caseload	295,010	143,283	4,076,889	1,117,048	842,684	485,409
Average monthly entries to, exits from, and transfers in and out of child poverty-related coverage groups
Entries	3.8	5.6	10.4	5.1	7.0	5.5
Exits	-3.2	-5.0	-10.8	-6.2	-5.5	-4.5
Transfers in (from other groups)	0.6	NA	11.2	4.8	2.8	0.3
Transfers out (to other groups)	-0.8	NA	-9.2	-3.6	-2.3	-0.6
Net difference	0.4	0.6	1.8	0.2	2.0	0.7
Average monthly caseload	121,059	44,363	63,221	223,983	86,881	58,383
Source:  1995 State Medicaid Research File data.

The turnover rate among the child poverty-related group is shown in the lower panel of Table 12. We were particularly interested in the level of turnover among this group, since policymakers have expected its enrollment to grow as more children became eligible under phased-in coverage and as outreach efforts made more parents aware of the availability of this coverage. In addition to showing the rates of entrants and exits for this group, Table 12 shows the average percentage of children who transferred in and out of poverty-related coverage from other eligibility groups each month. We expected that many children would be exiting the poverty-related groups, since these children usually are from higher-income families (making their Medicaid eligibility more tenuous) and some children would always be aging out of poverty-related coverage. In addition, during 1995, the economy was beginning to improve in many states, which might have caused a reduction in the number of children eligible for poverty-related coverage.

The average monthly rate of entrants into the poverty-related child group was greater than the exit rate in all the study states but California and Florida. In addition, in three of the states, the proportion transferring in (from other Medicaid groups) exceeded the proportion transferring out. On net, as reported above, enrollment in the child poverty-related group grew in all the study states during 1995.

In some respects, the sheer number of children and parents who exited Medicaid (or turned over) during 1995 may be just as important a measure as the overall caseload declines across the study states. Across the five study states, monthly Medicaid enrollment dropped about 180,000 from January to December of 1995. However, nearly 1.5 million children and almost a million adults were enrolled at some point during 1995, but not in December (Table 13). In California alone, more

TABLE 13
NUMBER OF CHILDREN AND ADULTS IN STUDY STATES
WHO DISENROLLED DURING 1995

	Alabama	Arkansas	California	Florida	Michigan	New Jersey	Total
Children	69,390	56,209	774,176	339,827	159,632	95,585	1,494,819
Adults	45,482	29,700	489,971	198,000	108,854	61,822	933,829
Total	114,872	85,909	1,264,147	537,827	268,486	157,407	2,428,648
Source: State Medicaid Research Files

TABLE 4 - California
Enrollment Pattern by Enrollment Group

Principal Enrollment Group	Single Coverage Gap	Other Enrolled in December	Other Not Enrolled in December
AFDC Cash	107,640	12,621	7,443	127,704	afdc
Transitional AFDC	3,509	232	899	4,640
EvM	10,975	829	2,906	14,710
Poverty-Related Eligibles	11,184	5,844	6,466	23,494	pov
Medically Needy	82,837	34,940	37,986	155,763	mn
Other	25,728	9,191	9,428	44,347
Disabled	1,611	384	213	2,208
All Persons	243,484	64,041	65,341	372,866
				65,905	other
				221,668

TABLE 4 - Florida
Enrollment Pattern by Enrollment Group -Number of Enrollees

Principal Enrollment Group	Single Coverage Gap	Other Enrolled in December	Other Not Enrolled in December
AFDC Cash	59,692	14,319	13,173	87,184	afdc
Transitional AFDC	5,279	707	2,063	8,049
Poverty-Related Eligibles	31,158	7,048	6,494	44,700	pov
Medically Needy	1,681	2,383	5,551	9,615	mn
Other	8,265	3,251	4,267	15,783
Disabled	1,098	298	124	1,520
				25,352	other
All Persons	107,173	28,006	31,672

than 750,000 children left Medicaid during the year. As a point of comparison, California’s target for SCHIP enrollment in 1998 was 200,000 children. The state might be well served to focus on children exiting Medicaid each month as part of its SCHIP outreach effort. Nationwide, the SCHIP target is 2 million children.

As a final part of our analysis, we looked at the extent to which churning accounted for caseload turnover. Churning refers to persons whose enrollment in Medicaid is interrupted during the year-- they were on, then off, then back on, and so forth. Managed care organizations complain about the loss of revenue and administrative expense associated with churning. We excluded the medically needy group from our churning analysis, since there is often some interruption in enrollment as a result of the spend-down requirements.

Florida had the highest level of churning during 1995, with 10.2 percent of its overall caseload experiencing interruptions in coverage during the year (Table 14); New Jersey had the lowest level, involving only 3.9 percent of its caseload. In Alabama, the level of churning was about the same across eligibility groups. In the other states, however, the poverty-related group had a higher level of churning than either the AFDC or the residual “other” group. This was particularly true in California, where the churning rate for the poverty-related group was 14.4 percent, which far exceeded the rates for the AFDC and the other group. We are unclear about why this pattern would occur.

Discussion

Study findings add to the body of evidence--that declines in the welfare caseload which began in 1995 are likely to have a noticeable effect on state Medicaid programs in the future, in terms of overall enrollment, caseload mix, and per capita expenditure levels. The findings also point to problems of continuity in Medicaid enrollment, which may be contributing to the lack of insurance among low-income children and their parents. Medicaid enrollment patterns that are cause for concern include:

TABLE 14
DISTRIBUTION BY ELIGIBLITY GROUP OF ENROLLEES
WITH MEDICAID ELIGIBILITY INTERRUPTIONS DURING 1995

	Alabama	Arkansas	California	Florida	Michigan	New Jersey
Ever Enrolled
AFDC cash	158,335	94,971	3,106,304	910,522	692,297	384,861
Poverty-related	205,398	87,801	163,561	400,298	182,658	112,083
Other	39,991	22,067	887,585	266,375	95,500	136,060
All persons	403,724	204,839	4,157,450	1,577,195	970,455	633,004
Enrollees with Eligibility Interruptions During 1995 "Churners"
AFDC cash	9,567	10,284	127,704	87,184	47,502	14,664
Poverty-related	12,476	6,014	23,494	44,700	14,935	4,868
Other	2,423	1,793	65,905	25,352	5,512	5,433
All persons	24,466	18,091	217,103	157,236	67,949	24,965
Percent Churners
AFDC cash	6.0	10.8	4.1	9.6	6.9	3.8
Poverty-related	6.1	6.8	14.4	11.2	8.2	4.3
Other	6.1	8.1	7.4	9.5	5.8	4.0
All persons	6.1	8.8	5.2	10.0	7.0	3.9
Source: State Medicaid Research Files Note: Medically needy are excluded from this analysis

TABLE 4 - California
Enrollment Pattern by Enrollment Group

Principal Enrollment Group	Single Coverage Gap	Other Enrolled in December	Other Not Enrolled in December
AFDC Cash	107,640	12,621	7,443	127,704	afdc
Transitional AFDC	3,509	232	899	4,640
EvM	10,975	829	2,906	14,710
Poverty-Related Eligibles	11,184	5,844	6,466	23,494	pov
Medically Needy	82,837	34,940	37,986	155,763	mn
Other	25,728	9,191	9,428	44,347
Disabled	1,611	384	213	2,208
All Persons	243,484	64,041	65,341	372,866
				65,905	other
				221,668

TABLE 4 - Florida
Enrollment Pattern by Enrollment Group -Number of Enrollees

Principal Enrollment Group	Single Coverage Gap	Other Enrolled in December	Other Not Enrolled in December
AFDC Cash	59,692	14,319	13,173	87,184	afdc
Transitional AFDC	5,279	707	2,063	8,049
Poverty-Related Eligibles	31,158	7,048	6,494	44,700	pov
Medically Needy	1,681	2,383	5,551	9,615	mn
Other	8,265	3,251	4,267	15,783
Disabled	1,098	298	124	1,520
				25,352	other
All Persons	107,173	28,006	31,672

· Declines in state welfare caseloads are associated with declines in Medicaid enrollment, particularly for parents. Every state showed a decrease in Medicaid enrollment during 1995, among family members qualifying for coverage through welfare. Some of the states were able to increase enrollment through the nonwelfare eligibility groups by making greater use of the child poverty-related, transitional assistance, and medically needy groups. Nevertheless, all the states had a net decline in the Medicaid enrollment of children and their parents during the study year. Since state welfare caseloads declined at much greater rates after 1995, there is concern that Medicaid enrollment will drop even more, and that Medicaid reductions may be contributing to the continuing problem of the uninsured. However, the SCHIP expansions for children (and the associated outreach) provide some hope that these Medicaid enrollment declines for children eventually will be reversed, and that fewer low-income children will be uninsured.

The situation for parents is less hopeful unless states make major changes for this group as well. To start, states could review their procedures for making sure that families formally or informally diverted from applying for welfare are encouraged to apply for Medicaid. As we discuss further below, states also need to give greater attention to the process for making sure eligible families leaving welfare continue their enrollment in Medicaid. Finally, states could make greater use of the provisions in the PRWORA legislation that allow expansion of Medicaid coverage provisions for low-income families. A few states--including Rhode Island, the District of Columbia and Connecticut--have used the PRWORA provisions to change their requirements for the poorest families. In these states, Medicaid coverage for parents will be more equivalent to the coverage available to children through the poverty-related expansion groups. Even modest expansions in coverage for parents may also have the added benefit of improved participation rates among children.

· Many welfare leavers are not staying on Medicaid and are at risk of becoming uninsured. At least half the children and adults leaving welfare in the study states also left Medicaid. Work by Short et al. (1998) and Garret and Holahan (1999) suggests that a significant number of welfare leavers become uninsured. This problem existed prior to PRWORA, but the rapid declines in welfare caseloads after 1995 may have exacerbated the effect on Medicaid enrollment levels. State officials report that many families leave welfare without providing the information needed to redetermine their continued eligibility for Medicaid. Some families leaving welfare may not need Medicaid, since they will be able to secure private insurance through employment; but others who become uninsured may mistakenly believe that they no longer qualify for Medicaid. It is also likely that some welfare leavers perceive the effort associated with maintaining Medicaid coverage not worth the benefit; some families have indicated that they know they can reapply for Medicaid if they need to.

· Medicaid programs appear to disproportionately lose low-cost welfare leavers while continuing to cover those with higher costs. Study results show that most adults and children who left welfare and Medicaid had lower average Medicaid expenditures compared to those who stayed on Medicaid. Findings also showed that adults and children with any amount of Medicaid expenditures were more likely to maintain Medicaid coverage upon leaving welfare. In addition, at least among children, as monthly expenditures increased, so did the probability of remaining enrolled in Medicaid. These results are similar to the findings of Garret and Holahan (1999), which suggest that health status is an important determinant of maintaining Medicaid coverage upon leaving welfare.

Welfare leavers who also leave Medicaid may be healthier and use fewer medical services prior to welfare exit. However, if these individuals leave welfare and become uninsured, they may jeopardize their future health status if they lack ready access to care and delay needed services. There is also concern for children in particular, that lack of insurance may keep them from obtaining appropriate preventive services.

Another issue involves managed care capitation rates. If Medicaid programs experience disproportionately higher rates of disenrollment of healthy, low-cost welfare leavers, the casemix of the Medicaid population may be affected, especially as it relates to managed care plans. States may find that capitated payment rates for managed care need to be adjusted upward if healthy welfare leavers continue to disenroll from Medicaid at greater rates. Alternatively, states will need to implement strategies that enable them to maintain Medicaid coverage for all welfare leavers who are Medicaid-eligible.

· States are experiencing considerable turnover in their Medicaid caseloads. Turnover in Medicaid enrollment was especially high for adults, ranging from 26 to 40 percent across the study states. This is consistent with the finding of brief Medicaid spells for adults by Short et al. (1998) using SIPP data. The turnover rates for children were considerably lower, but there was still an enormous number of children who left the Medicaid program during 1995. Recent research has shown that many children who are uninsured were previously enrolled in Medicaid. Czajka (1999), using 1992-1994 SIPP data, found that nearly 40 percent of the new spells of uninsurance among children were preceded by Medicaid. He also found that just over half of children who left Medicaid were without insurance the next month, and more than half of these still appeared to be eligible. Finally, some state SCHIP programs are finding that many of the children they locate through their outreach efforts were previously enrolled in Medicaid (Shenkman et al. 1998).

Across the five study states, almost 1 million adults and nearly 1.5 million children were enrolled at some point during 1995, but had disenrolled by the end of the year. This level of turnover may well be contributing to the sustained problem of the uninsured. The Balanced Budget Act of 1997 gave state Medicaid programs a new option to address the problem of turnover, at least for children. States can now elect to guarantee child Medicaid enrollment for up to 12 months (after each eligibility redetermination). It will be important to track the experience of states who choose this option to see what impact it has on enrollment and expenditure levels.

Congress and the states continue to be concerned about the high rate of uninsured, in spite of the robust economy and continued Medicaid expansions. The results from this study suggest that the dynamics of Medicaid enrollment may be at least part of the problem. If state Medicaid programs could reduce the number of children and parents leaving welfare without continuing on Medicaid, that would be a step in the right direction. In addition, greater attention needs to be given to inappropriate turnover among children and adults qualifying for Medicaid through the nonwelfare groups. Although states may continue to make policy changes to expand Medicaid coverage provisions in an effort to reduce the number of uninsured, these changes are not likely to be effective unless steps are also taken to improve individual continuity in Medicaid coverage.

(1)The “adult” group in Medicaid generally covers the parents and other caretaker relatives of children.

(2)Noncash eligibility groups include families eligible for cash assistance but not participating; families eligible for less than $10 in AFDC benefits (who would not get AFDC, but only Medicaid benefits); families who would qualify for AFDC if AFDC were as broad as federally allowed; families who would qualify for AFDC if child care costs were paid from earnings; persons who would be eligible if they were not in medical institutions; and those who were not lawful, permanent residents (who qualified only for emergency services).

(3)We did not expect the AFDC declines reported in the SMRF data to match exactly the declines reported in the AFDC program data for the same period. State Medicaid programs report that there is often some lag before AFDC disenrollments are reflected in Medicaid data. In addition, Medicaid data include retroactive adjustments for eligibility that do not occur with AFDC program data.

(4)Beginning in 1990, a class-action court decision (Edwards v. Kizer) required California to allow families discontinued under cash assistance to remain on Medicaid until their eligibility under other coverage groups was redetermined. At the beginning of 1995, California had a large backlog of families in this group, due to staff shortages and delays; by the end of 1995, this backlog had been considerably reduced.

(5)Because the Medicaid system in California is county-based and county identifiers are available in the SMRF data, we initially attempted to identify managed care and nonmanaged care counties. We found this definition unsatisfactory and the resulting sample of persons in nonmanaged care counties had expenditure distributions that were not similar enough to the fee-for-service samples from the other states to make us confident that our California sample was adequately defined.

(6)Monthly Medicaid expenditures were based on total Medicaid expenditures for calendar year 1995 adjusted for the number of months a person was enrolled in 1995.

(7)As an example, among adults in Michigan the first quintile of monthly expenditures ranged from less than one dollar to $20.78, the second from $20.83 to $46.58, the third from $46.63 to $97.08, the fourth from $97.11 to $264.33, and the fifth quintile ranged from $264.42 to $17,042.50. Among children the quintiles were: (1) less than one dollar to $9.58; (2) $9.60 to $19.00; (3) $19.08 to $33.00; (4) $33.08 to $65.83; and (5) $65.89 to $19,357.92.

(8)The SMRF 1995 enrollment data are both right- and left-censored. We do not have information on Medicaid enrollment prior to January 1995 and subsequent to December 1995. The length of enrollment data relate only to the duration of enrollment during 1995.

References

Czajka, John L. “Analysis of Children’s Health Insurance Patterns: Findings from the SIPP.” Washington, DC: Mathematica Policy Research, Inc., 1999.

Dubay, Lisa, and Genevieve Kenney. “Effects of Medicaid Expansions on Insurance Coverage of Children.” Future of Children, vol. 6, no. 1, 1996, pp. 152-61.

Ellwood, David T., and E. Kathleen Adams. “Medicaid Mysteries: Transitional Benefits, Medicaid Coverage and Welfare Exits.” Health Care Financing Review 1990 Annual Supplement, December 1990.

Ellwood, Marilyn, and Leighton Ku. “Welfare and Immigration Reforms: Unintended Side Effects for Medicaid.” Health Affairs, vol. 17, no. 3, May/June, 1998, pp. 137-151.

Ellwood, Marilyn, and Kimball Lewis. On and Off Medicaid: Enrollment Patterns for California and Florida in 1995. Washington, DC: The Urban Institute, 1999.

Garrett, Bowen, and John Holahan. “Health Insurance Coverage after Welfare.” Health Affairs, vol. 19, no. 1, January/February, 2000a, pp. 175-184.

Garrett, Bowen, and John Holahan. “Do Welfare Declines Make the Risk Pool Sicker?” Draft Working Paper. Washington, DC: the Urban Institute, 2000b.

Ku, Leighton, and Brian Bruen. The Continuing Decline in Medicaid Coverage. Washington, DC: The Urban Institute, 1999.

Perry, Michael J., Evan Stark, and R.Burciaga Valdez. Barriers to Medi-Cal Enrollment and Ideas for Improving Enrollment: Findings from Eight Focus Groups in California with Parents of Potentially Eligible Children. Menlo Park, CA: The Henry J. Kaiser Family Foundation, 1998.

Selden, Thomas M.; Jessica S. Banthin; and Joel W. Cohen. “Medicaid’s Problem Children: Eligible But Not Enrolled.” Health Affairs vol. 17, no. 3, May/June, 1998. pp. 192-200.

Shenkman, Elizabeth, D. Wegener, and C. Bono. “KidCare: The Impact on Medicaid Eligibles in the Healthy Kids Program.” Tallahassee, FL: Institute for Child Health Policy, 1998.

Shuptrine, Sara, Vicki Grant, and Genny McKenzie. Southern Regional Initiative To Improve Access to Benefits for Low Income Families with Children. Columbia, S.C.: Southern Institute on Children and Families, 1998.

Smith, Vernon, Robert Lovell, Karin Peterson, and Mary Jo O’Brien. The Dynamics of Current Medicaid Enrollment Changes. Washington, DC: Kaiser Commission on Medicaid and the Uninsured, 1998.

Appendix

APPENDIX TABLE 1
LOGISTIC REGRESSION OF THE PROBABILITY
OF MAINTAINING MEDICAID COVERAGE
WHEN ADULTS LEAVE AFDC
(Adults in HMOs Excluded)

Characteristic	Alabama	Florida	Michigan	New Jersey
	Estimated Odds Ratio a
Female	1.68***	1.04	0.88***	1.26***
Race/Ethnicity:b
Black	1.10	1.82***	0.50***	0.62***
Hispanic	1.62	0.80***	0.71***	0.67***
Ages as of 12/31/95:c
15 through 20	1.25***	0.80***	0.96	1.18*
21 through 24	1.57***	0.82***	1.02	1.39***
25 through 30	1.34***	0.77***	1.06	1.23***
31 through 34	1.37***	0.95	1.17***	1.19***
Average Monthly Expenditures:d
First Quintile	2.02***	2.20***	1.95***	3.47***
Second Quintile	2.21***	2.92***	1.99***	3.50***
Third Quintile	2.57***	3.64***	2.02***	3.36***
Fourth Quintile	2.81***	4.05***	2.19***	3.21***
Fifth Quintile	2.63***	3.45**	1.56***	2.40***
Black
Average Monthly Expenditures:d
First Quintile	1.41	0.93	1.54***	0.93
Second Quintile	1.28	0.92	1.67***	1.16
Third Quintile	1.38	0.75*	1.51***	1.05
Fourth Quintile	1.52*	0.93	1.54***	1.06
Fifth Quintile	1.29	0.64***	1.57***	1.02
Number of Observations	6,957	12,626	20,973	9,279
SOURCE:  1995 State Medicaid Research File, Summary Person File. aThe estimated odds ratio is the estimate of the relationship between the characteristic and p/(1-p) where p is the probability that the person is enrolled in Medicaid three months after leaving AFDC, and (1-p) is the probability a person is disenrolled from Medicaid three months after leaving AFDC. bCompared to whites. cCompared to adults aged 35 through 64. dCompared to no expenditures. *Significantly different from zero at the .10 level, two-tailed test. **Significantly different from zero at the .05 level, two-tailed test. ***Significantly different from zero at the .01 level, two-tailed test.

APPENDIX TABLE 2
LOGISTIC REGRESSION OF THE PROBABILITY
OF MAINTAINING MEDICAID COVERAGE
WHEN CHILDREN LEAVE AFDC
(Children in HMOs Excluded)

	Alabama	Florida	Michigan	New Jersey
	Estimated Odds Ratio a
	1.02	0.95*	1.03	0.96
Race/Ethnicity:b
	2.54***	1.96***	0.98	0.91
	1.16	0.80***	0.75***	0.68***
Ages as of 12/31/1995:c
	0.94	0.91***	1.14***	1.09**
15 through 20	0.26***	0.55***	0.52***	1.03
Average Monthly Expenditures:d
	4.26***	3.21***	3.03***	3.76***
Second Quintile	5.15***	3.83***	3.44***	3.67***
Third Quintile	4.94***	4.49***	3.68***	4.14***
Fourth Quintile	6.84***	4.91***	3.67***	4.37***
Fifth Quintile	9.16***	5.53***	4.22***	4.67***
Average Monthly Expenditures:d
First Quintile	0.78	0.95	0.82**	0.93
Second Quintile	0.64***	1.00	0.86	1.16
Third Quintile	0.79	0.76***	0.84*	1.05
Fourth Quintile	0.64***	0.72***	0.82*	1.06
Fifth Quintile	0.61***	0.77***	0.82*	1.02
Number of Observations	14,613	23,382	31,937	16,956
SOURCE:  1995 State Medicaid Research File, Summary Person File. aThe estimated odds ratio is the estimate of the relationship between the characteristic and p/(1-p) where p is the probability that the person is enrolled in Medicaid three months after leaving AFDC and (1-p) is the probability a person is disenrolled from Medicaid three months after leaving AFDC. bCompared to whites. cCompared to children under six years of age. dCompared to no expenditures. *Significantly different from zero at the .10 level, two-tailed test. **Significantly different from zero at the .05 level, two-tailed test. ***Significantly different from zero at the .01 level, two-tailed test.