Using Past Income Data to Verify Current Medicaid Eligibility Stan Dorn Matthew Buettgens Habib Moody Christopher Hildebrand The Urban Institute Health Policy Center 2100 M St. NW Washington, DC 20037 October 2013 Copyright © October 2013. The Urban Institute. The views expressed are those of the authors and should not be attributed to the Urban Institute, its trustees, or its funders. Permission is granted for reproduction of this file, with attribution to the Urban Institute. Executive Summary Context The Patient Protection and Affordable Care Act (ACA) changes the rules for verifying financial eligibility for Medicaid. Beginning in January 2014, such verification must occur through automated data matches whenever possible, rather than by requiring consumers to submit paperwork that is manually evaluated, at public expense. This new data-driven approach was adopted to streamline enrollment and retention, increase eligible consumers’ participation, cut administrative costs, and prevent errors. But when, exactly, do data matches from records showing earlier income verify consumers’ current financial eligibility for Medicaid? Put differently, what business rules* can states program into their eligibility systems that will identify the precise circumstances under which data matches verify financial eligibility, without any need for documentation from the consumer or manual intervention by caseworkers? To help states develop such rules, we analyze data from the 2008 Survey of Income and Program Participation (SIPP), which tracks the changing circumstances of households over time. We explore the potential usefulness of federal and state tax, employment, and wage records in states that implement the ACA’s Medicaid expansion to 138 percent of the federal poverty level (FPL). We estimate, among people with specified prior incomes, wages, and employment histories, the proportion who will qualify for Medicaid under rules in effect in 2014 and later. We use those estimates to develop a set of business rules that would let states verify financial eligibility, without seeking documentation from consumers, both when consumers submit initial Medicaid applications and when Medicaid enrollees’ eligibility is redetermined. These business rules consider the changing information available to states throughout each calendar year. During the first few months of any given year, states’ most important source of financial information is likely to be tax records showing prior-year income. In later months, this information is supplemented by reports about whether particular consumers began new jobs during the current year. By the latter part of the year, states can also access wage records from at least the first calendar quarter of the current year. SIPP lets us establish the probability of current Medicaid eligibility for people with certain levels of past income. But federal guidance on this issue has not been framed in terms of probabilities. For our analysis, we assume that the Centers for Medicare & Medicaid Services (CMS) will give states significant leeway in setting applicable probability thresholds, and we base our business rules on two specific thresholds—one for initial applications, and the other for renewals of existing beneficiaries—that should be well within the bounds of states’ permitted flexibility. Initial applications When someone submits an initial application, ACA regulations provide that: (a) the applicant’s attestations are verified by “reasonably compatible” data matches; (b) states have considerable flexibility to define such compatibility; and (c) if both the attestation and the data show income below 138 percent FPL, the applicant is financially eligible. Based on these principles, we assume that if data matches show prior income below 138 percent FPL, and they establish more * Business rules are the operating instructions for computer-driven eligibility determination in the same way that eligibility manuals have traditionally provided operating instructions for caseworker-driven eligibility determination. i than a 50 percent likelihood of current income at Medicaid levels, a state can define such data as “reasonably compatible” with an applicant’s attestation of Medicaid-level income. To err on the conservative side, we look for data matches that verify attestations of financial eligibility by showing prior income at Medicaid levels and a 70 percent or greater likelihood of current financial eligibility. In 2014, a state that bases initial Medicaid eligibility on monthly income could satisfy those criteria with the business rules shown in Table ES-1. Table ES-1. Possible business rules for initial applications in 2014: applicants whose attestations of current Medicaid-level income are verified by data matches Time frame Data that verify an attestation of current financial eligibility January through April 2014 May through August 2014 2013 income no higher than 138 percent FPL Either: (a) 2013 income no higher than 138 percent FPL and no new job between 1/1/2014 and 5/1/2014; or (b) 2013 income no higher than 100 percent FPL and a new job begun between 1/1/2014 and 5/1/2014 Sept. 2013 income at or below 138 through percent FPL plus quarterly wages, Dec. 2014 earlier during 2014, at or below 80 percent FPL Based on the data, the likelihood of current monthly income at Medicaid levels 85% Among all Medicaideligible consumers, the percentage potentially verified by the rule 79% (a) 82% 73% (b) 70% 85% 55% Source: SIPP 2008. For example, if someone applying in January through April 2014 attests to income at Medicaid levels, a data match showing 2013 income at or below 138 percent FPL would establish an 85 percent likelihood of Medicaid-level current monthly income. It would therefore be reasonably compatible with the attestation and verify it, eliminating the need for the applicant to produce paperwork. Such a rule could verify attestations for 79 percent of eligible consumers. Renewals When enrollees are reaching the end of their coverage periods, ACA regulations provide that the Medicaid program “must make a redetermination of eligibility without requiring information from the individual if able to do so based on reliable information.” When such information shows continuing eligibility, the beneficiary is sent a notice explaining the basis of the decision to continue coverage. The notice makes clear that the beneficiary is legally required to correct any errors. If no corrections are received, coverage continues. This process is termed “administrative renewal.” CMS has not defined the regulatory phrase “reliable information” in terms of the likelihood of current eligibility that is needed for administrative renewal. For purposes of this analysis, we assume that states have the flexibility to apply administrative renewal if reliable data sources ii establish at least an 80 percent likelihood of continuing eligibility. Under such circumstances, administrative renewal is likely to improve the overall accuracy of eligibility determination by reducing the number of eligible people who lose coverage for procedural reasons. To err once again on the conservative side, we look for data matches that establish at least a 90 percent likelihood of current financial eligibility. In 2014, a state that uses such a 90 percent threshold and redetermines Medicaid eligibility based on projected annual income during the current year could use the business rules shown in Table ES-2. Table ES-2. Possible business rules for renewal in 2014: beneficiaries whose eligibility is administratively renewed Time frame Data match that triggers administrative renewal January through April 2014 May through August 2014 Sept. through Dec. 2014 2013 income no higher than 120 percent FPL 2013 income no higher than 120 percent FPL and no new job begun between 1/1/2014 and 5/1/2014 2013 income at or below 138 percent FPL and quarterly wages, earlier during 2014, at or below 80 percent FPL Based on the data match, the likelihood of 2014 annual income at Medicaid levels 90% Among all Medicaid-eligible consumers, the percentage potentially renewed administratively by the rule 92% 63% 98% 60% 71% Source: SIPP 2008. Under these rules, if someone’s eligibility is being redetermined between January and April 2014, data showing 2013 income at or below 120 percent FPL would establish a 90 percent likelihood of 2014 income at Medicaid levels. It would therefore trigger administrative renewal, continuing coverage unless the beneficiary corrects the state’s notice. This rule could administratively renew 71 percent of eligible beneficiaries during those months. Conclusion States could craft different business rules, of course, based on different probability thresholds than those we used, different budget periods, and other factors. To aid the development of such alternative rules, both the body of the report and appendices include detailed findings about the relationship between the past and present circumstances of low-income people. We also provide an example of alternative business rules that are premised on different state policy choices. Regardless of a state’s rules, many consumers will need to document financial eligibility for Medicaid. Nevertheless, for numerous eligible consumers, data matches will be sufficient under the ACA to verify eligibility in ways that cut red tape for families, save administrative resources for the state and federal government, increase participation among eligible consumers, safeguard program integrity, and reduce the number of incorrect outcomes. iii Contents Introduction ..................................................................................................................................... 1 The level of certainty required for data matches to verify eligibility ............................................. 4 Initial application ................................................................................................................ 4 Renewal............................................................................................................................... 5 Budget period for current eligibility: monthly vs. annual income ...................................... 7 Income data accessible to state Medicaid programs ....................................................................... 7 Prior-year income information ............................................................................................ 7 Current-year income data .................................................................................................... 9 Initial applications: How Medicaid programs can use past income data to verify attestations of financial eligibility .......................................................................................................................... 9 January through April: Verifying financial eligibility when only prior-year income data are available ...................................................................................................................... 10 May through August: Verifying financial eligibility when significant new hires data become available ............................................................................................................... 11 September through December: Verifying financial eligibility when wage data become available from earlier quarters .......................................................................................... 13 Redeterminations: How Medicaid programs can use past income data to trigger administrative renewal .......................................................................................................................................... 15 January through April: Verifying financial eligibility when only prior-year income data are available ...................................................................................................................... 15 May through August: Verifying financial eligibility when significant new hires data become available ............................................................................................................... 16 September through December: Verifying financial eligibility when wage data become available from earlier quarters .......................................................................................... 17 Putting it into practice: How business rules for data-based verification could operate in 2014 ... 19 Conclusion .................................................................................................................................... 23 Additional Tables .......................................................................................................................... 24 Appendix A: California findings .................................................................................................. 33 Appendix B. Methodology............................................................................................................ 34 Appendix C. Additional results..................................................................................................... 36 About the Authors and Acknowledgments ................................................................................... 39 About the Urban Institute.............................................................................................................. 40 About the California HealthCare Foundation ............................................................................... 40 Notes ............................................................................................................................................. 40 iv Using past income data to establish current Medicaid eligibility Introduction The Patient Protection and Affordable Care Act (ACA) makes several important changes to the Medicaid program. The best known change involves who is eligible. Beginning in 2014, the ACA expands Medicaid to include all citizens and qualified immigrants with modified adjusted gross income (MAGI) at or below 138 percent of the federal poverty level (FPL)1—an expansion that the U.S. Supreme Court effectively rendered optional for states.2 As of September 30, 2013, 25 states were moving forward and 26 states were not moving forward with expansion.3 But the ACA also changes how eligibility is determined, and this change is not optional. If an applicant’s statement on a Medicaid form, sworn under penalty of perjury, is “reasonably compatible” with data from reliable sources (like income tax and quarterly wage records), eligibility is verified without asking the applicant to furnish documentation. Likewise, if such data show that a beneficiary continues to qualify at the end of his or her enrollment period, coverage continues, without any need for the beneficiary to complete and return renewal forms.4 This new approach to determining eligibility uses 21st-century information technology (IT) in pursuit of several goals: streamlining enrollment and renewal, thereby increasing participation levels among eligible consumers; reducing state and federal administrative costs; and preventing eligibility errors. The Centers for Medicare & Medicaid Services (CMS) have offered states enhanced federal funding to support the IT development needed to realize this vision.5 The federal government will also provide multiple sources of information through a federal data hub, including tax data from the Internal Revenue Service (IRS), citizenship and benefits information from the Social Security Administration (SSA), immigration status data from the Department of Homeland Security, and information from other sources, including private vendors. Medicaid programs will supplement this information with other data, including from state-specific sources. Under the ACA, states retain the option to base financial eligibility determinations entirely on attestations. However, as in the past, most states expect to require income verification. Such verification must be obtained either at the time of application or soon thereafter. This paper focuses on one question state officials have raised about how to implement the ACA’s new approach —namely, when do data matches from earlier income records adequately verify that current income is low enough to qualify for Medicaid? That question has a specific operational context. The ACA’s move to data-driven eligibility envisions that, whenever possible, states will determine eligibility through the use of “rules engines”—computer-operated systems that use clearly articulated business rules that, among other things, define the precise circumstances when data matches and consumer-furnished information suffice to establish eligibility. Business rules are the operating instructions for computer-driven eligibility determination in the same way that eligibility manuals have traditionally provided operating instructions for caseworker-driven eligibility determination. A key goal of this paper is to help states develop business rules they can program into their eligibility systems to identify the precise circumstances under which data matches verify 1 financial eligibility, reducing the need for documentation from consumers and manual intervention by caseworkers. To accomplish this goal, we analyze data from the 2008 Survey of Income and Program Participation (SIPP). This “panel” of SIPP began tracking a group of respondents in 2008 and followed their changing circumstances for several years;6 the first full calendar year for which all respondents provided data was 2009. We estimate, among people who had specified levels of income during the prior calendar year, the proportion who would qualify for Medicaid during the current month or year. We do this by examining the relationship, in the SIPP, between annual income during 2009—which we define as “prior-year” income—and monthly or annual income in 2010—which we define as “current” monthly or annual income. We also estimate the effects of supplementing prior-year income data with information about quarterly wage records and new hires during the “current year”—that is, 2010 in the SIPP. We assume that states implement the ACA’s eligibility expansion to 138 percent FPL. This paper begins by analyzing the level of certainty required to rely on data matches in verifying eligibility, both for initial applications and for renewals. We then describe state Medicaid programs’ access to relevant income and earnings data. We explain our findings that show the relationship between prior-year income and current financial eligibility for Medicaid, estimating the proportion of Medicaid-eligible consumers whose current financial eligibility can be verified based on such records. We also assess the value added by information about quarterly wages and the start of new employment earlier during the current year. Appendices explore whether similar trends apply to the U.S. as a whole and to California, the nation’s most populous state; explain our methodology in more detail; and provide additional results. To present our results, we discuss in turn data matches that verify initial applications and those that verify renewals. Within each of these two sections, we proceed chronologically, taking state Medicaid agencies through the calendar year, from January through December. We begin with early months when nothing but prior-year income data are available; move to the middle of the year, when prior-year data are supplemented by reports indicating whether consumers began new jobs since January 1 of the current year; and conclude with latter part of the year, when all Medicaid programs have access to wage records from at least one quarter during the current year. In the final section of the paper we present two illustrative sets of business rules, which reflect varying policy goals and assumptions. We preface this analysis with several final introductory comments. First, our analysis focuses on SIPP respondents who reported various levels of income in 2009, identifying the proportion who subsequently reported income at or below 138 percent FPL during certain periods in 2010. This was a time of economic difficulty. During 2010, aggregate quarterly personal income was an average of 3.8 percent higher than during the same quarter in 2009.7 If future economic growth is more rapid than it was from 2009 to 2010, a higher proportion of consumers with low income during one year could be financially ineligible for Medicaid in the next. Second, the business rules we develop in this analysis treat all nonelderly consumers equally, without distinguishing between children and adults. State Medicaid programs could use separate business rules to verify financial eligibility for children and adults: because of the ACA’s maintenance-of-effort requirements, children will typically qualify for Medicaid at higher income thresholds than adults, and eligibility could thus be verified via data matching more frequently for children than for adults. However, our analysis combines children and adults. We 2 do so for several reasons: eligibility thresholds for children vary greatly by state, so children’s verification rules would need to differ as well; and SIPP does not include large enough samples to produce reliable state-level estimates throughout the country, so it does not permit developing state-specific rules. States in the future may be able to use administrative data to craft different verification rules for children and parents, but in the near-term, many states that cover both adults and children up to 138 percent FPL are likely to use the same verification procedures for all residents under age 65 whose eligibility is based on modified adjusted gross income (MAGI). Finally, our focus here is exclusively on the relationship between income tax data, quarterly wage records, and new hire reports, on the one hand, and current financial eligibility, on the other. But other data can also verify that particular consumers qualify for Medicaid. For example, earnings reports from private vendors can supplement publicly maintained income records to provide states with additional certainty about current eligibility. The Federal Data Hub will make information available from The Work Number, a subsidiary of Equifax that provides real-time payroll information about workers in more than 2,000 large firms, which may suffice to verify current monthly income levels for numerous consumers. As another example, CMS has made clear that receipt of public benefits can be used as part of a state’s verification plan.8 For example, 97 percent of beneficiaries of the Supplemental Nutritional Assistance Program (SNAP) will qualify for Medicaid under the ACA’s 2014 rules.9 Data showing receipt of SNAP is thus more than sufficient to verify a new applicant’s attestation of financial eligibility for Medicaid or to justify administrative renewal of a beneficiary already receiving Medicaid. Ultimately, Medicaid programs will operate with business rules that use a broad variety of data matches to verify eligibility. This report focuses on one subset of such rules—namely, those that use publicly-maintained records of past income, wages, and the start of new employment to verify current financial eligibility. It seeks to build part of what will become a much larger repertoire of rubrics specifying when data matches have eliminated the need for further documentation. 3 Data-driven eligibility in Oklahoma: a real-world example. The one state that, before October 2013, operated an ACA-style Medicaid application and renewal system—Oklahoma—takes applications on-line and uses a rules engine to provide real-time eligibility determinations, based on data matches whenever possible. Some data, such as citizenship verification from SSA, is obtained in real time, but other data sources, such as wage records from the state workforce agency, are queried after the application is submitted and eligibility has been initially determined based on the applicant’s attestations. The latter determinations are subject to Oklahoma’s right to revise its decision after accessing relevant data. If available data are inconsistent with or insufficient to support the applicants’ attestations, the state contacts the consumer for additional information. Data matches fully verify eligibility for about 55 percent of online applications, so slightly more than half of applicants do not have to provide any documentation to enroll. For renewals, 80 to 85 percent of beneficiaries qualify based on data matches, without any need for new documentation. Although the system has created some new administrative costs, net administrative savings amount to $1.5 million per year. These improvements occurred without any increase in the state’s federally measured error rate, according to state officials.10 The level of certainty required for data matches to verify eligibility Here, we describe the rules CMS has articulated for using data to verify eligibility at initial application and renewal. We also explore the implications of those rules for this paper’s analysis. Initial application The ACA’s final Medicaid regulations require that, beginning in 2014, if a sworn attestation on a Medicaid application form and data matches are “reasonably compatible” with one another, eligibility is determined accordingly, without seeking further verification.11 In guidance, CMS defines “reasonable compatibility” to mean that “[i]nformation is relatively consistent and does not vary significantly or in a way that is meaningful for eligibility.”12 For example: If both an attestation and data matches show MAGI “below the applicable income standard,”13 they are reasonably compatible. When an attestation and relevant data matches are not reasonably compatible, the state asks the applicant for more information, such as a “statement which reasonably explains the discrepancy” or documentation “to the extent electronic data are not available…”14 State Medicaid programs have significant flexibility to define the differences between attestation and data that are consistent with reasonable compatibility.15 Each state must provide CMS with a “verification plan” that includes, among other things, the state’s definition of reasonable compatibility.16 If a state determines eligibility for a particular consumer by correctly applying the rules specified in its Medicaid State Plan and verification plan, and those rules meet the requirements of federal law, the state decision is not classified as erroneous for purposes of federal payment error review. So long as the correct process was followed, given the information available to the Medicaid program at the time, the state is not penalized.17 CMS has issued no hard-and-fast rules that define the timeliness of income data, instead emphasizing the role played by state flexibility. For example, CMS explained that current-year wage records need not be examined if the state finds that prior-year income data furnish 4 sufficient verification because they are reasonably compatible with applicant attestations about current income.18 CMS’s instructions for state verification plans thus ask states to consider whether “the age of the data (up to date, 1 month old, 3 months old, a year old) [is] useful for determining current income,”19 without articulating any criteria for determining when data has become too old to be useful. Here, we define the relevance of financial data based on the probability it establishes of current Medicaid eligibility, taking into account both the age and the content of the data. In this paper, we estimate the probability of current financial eligibility established by data showing past levels of income, wages, and the start of new employment. Unfortunately for purposes of this analysis, CMS’s guidance has not yet defined verification standards in terms of probability. The plain meaning of “reasonable compatibility” suggests that, if a data match with income records makes current eligibility more likely than not—that is, if such records establish more than a 50 percent likelihood of current financial eligibility—a state could classify the data as reasonably compatible with an attestation of income low enough to qualify for Medicaid, particularly if the data shows prior income at Medicaid levels. Under such circumstances, both the data and the attestation would show income below 138 percent FPL, which fits the federal definition of reasonable compatibility. In this paper, we do not explore this possible outer boundary of permitted state flexibility. Instead, we develop business rules based on the conservative assumption that a Medicaid program can verify attestations of financial eligibility with data matches that: (a) show income low enough to qualify for Medicaid during the most recent period of time for which information is available; and (b) establish a 70 percent or greater likelihood of current income at Medicaid levels. Such matches, combined with an applicant’s sworn statement under penalty of perjury that current income falls below 138 percent FPL, could qualify the applicant for Medicaid without any need to provide pay stubs or other documentation that supplements the data. Renewal Administrative renewal In 2014 and beyond, ACA regulations provide that, if “reliable information” 20 demonstrates that a Medicaid beneficiary receiving coverage continues to qualify, the state must renew the individual’s Medicaid eligibility. 21 The beneficiary is sent a notice describing the basis of the decision and explaining that the beneficiary is legally required to correct any errors. If no corrections are received, coverage continues. This process is termed “administrative renewal.” When available data make beneficiary’s ongoing qualification for Medicaid appear uncertain or unlikely, the “default” under the ACA in case of consumer inaction is termination. The state sends the beneficiary a “pre-populated” form containing information known to the Medicaid program and asks for whatever additional information is required to determine eligibility.22 Coverage ends unless the beneficiary provides that information. The ACA thus changes the default from termination to continued coverage whenever “reliable information” shows ongoing eligibility. This feature of the ACA reflects the insights of behavioral economics about the powerful role that default arrangements play in shaping participation levels for both public and private benefits.23 5 In the past, eligible beneficiaries often lost Medicaid or CHIP by default. Failure to complete and return renewal forms has caused numerous procedural terminations, often because of confusion, lost paperwork, procrastination, or other behavioral factors.24 The ACA’s administrative renewal procedures address this problem by requiring Medicaid programs to “make a redetermination of eligibility without requiring information from the individual if able to do so based on reliable information.”25 But where do states draw the line? What flexibility do states have to define “reliable information?” How likely must continuing eligibility be for states to set the default at administrative renewal rather than termination? CMS has not specifically addressed this question, but several basic points are clear. First, it is appropriate to require the establishment of a higher probability of current eligibility with administrative renewal than in verifying applicant attestations. With administrative renewal, the state receives some information from the beneficiary’s failure to correct the state’s eligibility notice, but less than is provided by an applicant’s sworn attestation of income. Since the state receives less information, it is reasonable that the data match must be more conclusive to verify financial eligibility. Second, if it is applied to beneficiaries who are known to have at least an 80 percent likelihood of continued eligibility, administrative renewal is likely to increase the overall accuracy of eligibility redetermination. Under a broad range of assumptions, administrative renewal used with such beneficiaries prevents the mistaken termination of eligible consumers more often than it causes the mistaken renewal of ineligible consumers.26 Put differently, changing the default, in the event of consumer inaction, from termination to renewal makes sense when the vast majority of enrollees are eligible. We therefore assume that states could apply administrative renewal to beneficiaries who are known to have at least an 80 percent probability of continuing eligibility. However, as with verification of applicants’ financial eligibility, our development of business rules below does not press the outer bounds of permitted state discretion. Rather, we look for rules that will trigger administrative renewal based on data showing a 90 percent or greater likelihood of ongoing eligibility. This approach is broadly consistent with CMS’s authorization of states to enroll some or all of their uninsured SNAP recipients into Medicaid, even though, in some states, the proportion of SNAP recipients who are financially eligible for Medicaid can be low as 93 percent for all recipients and 89 percent for adults.27 6 Streamlined renewals in Louisiana: a real-world example. Louisiana’s renewal policies for Medicaid and CHIP children illustrate the potential offered by administrative renewal under the ACA. When children’s eligibility periods are drawing to a close, the Medicaid system gathers data from multiple sources. If they show that continued eligibility is reasonably certain, coverage is renewed without requesting any information from the family. Under this policy: 75 percent of renewals take place without requiring information from beneficiaries, and fewer than 1 percent of children are terminated for procedural reasons.28 More than 95 percent of Medicaid and CHIP children continue receiving coverage.29 The state’s most recent eligibility error rate was 0.3 percent—one-tenth the national average. Only one case in Louisiana’s entire payment error sample had an eligibility error. 30 Manual renewal Even if administrative renewal is not available for a particular enrollee, data matches can still expedite redetermination of eligibility. CMS’s final Medicaid regulations specify that, at renewal, a Medicaid program “may request from beneficiaries only the information needed to renew eligibility.”31 The regulations further provide that the same “reasonable compatibility” standards that are used for initial applications also govern manual renewals.32 Accordingly, if continued eligibility is not sufficiently certain for administrative renewal, but data matches with income information meet the state’s definition of reasonable compatibility, a beneficiary should be able to renew coverage by simply attesting, under penalty of perjury, that current income does not exceed 138 percent FPL. Medicaid regulations make clear that consumers can provide such renewal attestations by phone, on-line, by mail, or in person.33 Budget period for current eligibility: monthly vs. annual income The ACA’s Medicaid regulations base financial eligibility for applicants on their current monthly income. At renewal, states have the option to base eligibility on either current monthly income or projected annual income for the remainder of the calendar year.34 However, CMS officials have indicated that states might be able to determine initial financial eligibility based on projected annual income or average rather than current monthly income.35 The use of annual or average monthly budget periods can reduce the instability created by fluctuations in month-to-month income. Also, by increasing a state’s ability to predict ongoing eligibility, basing eligibility on projected annual income or average monthly income may reduce a state’s exposure to federal findings of qualify control errors. Based on these considerations, we assume for the purposes of this analysis that a state (a) determines the initial eligibility of new applicants based on current monthly income but then (b) uses projected annual income to redetermine eligibility of existing beneficiaries. Income data accessible to state Medicaid programs Prior-year income information States will have access to information about prior-year income, as recorded on federal income tax returns, in several ways: 7 State Medicaid programs can access the ACA’s new federal data hub, which will provide the most recent information available from the Internal Revenue Service (IRS) about household size and MAGI. States may also be able to obtain information about prior-year income tax returns through systems that were in place before ACA enactment. These include the Income Eligibility Verification System, created by Social Security Act Section 1137,36 and the Social Security Administration’s Beneficiary & Earnings Data Exchange.37 IRS data are available for numerous Medicaid-eligible people. Most low-income households file federal income tax returns, often in order to claim the Earned Income Tax Credit (EITC) and Child Tax Credit or to obtain refunds of withheld wages. Even among the uninsured, 86 percent file federal income tax returns, including 75 percent of those with incomes below poverty.38 Whether through the federal hub or pre-ACA sources, one year’s tax return information should be available to state Medicaid programs soon after the next year begins, since most low-income households file their federal income tax returns well before April 15. Among returns claiming the EITC, for example, more than four in five (81 percent) are filed by the end of March (Figure 1). Figure 1. Among federal income tax returns that claimed an EITC for 2006, the proportion that were filed by various dates in 2007 48% 57% 64% 69% 73% 76% 79% 81% 83% 86% 95% 95% 96% 99% 100% 32% 2-Feb 9-Feb 16-Feb 23-Feb 2-Mar 9-Mar 16-Mar23-Mar30-Mar 6-Apr 13-Apr 20-Apr 27-Apr 4-May 24-Aug Oct. 26 Source: Authors’ calculation, IRS, Tax Year 2006 Taxpayer Usage Study. Note: Calculation assumes that all applicable returns were filed by October 26. “EITC” refers to the Earned Income Tax Credit. IRS officials have stated that, through the federal data hub, tax return information will be available to verify eligibility for insurance affordability programs (IAP) under the ACA as soon as such information has been “posted” within the IRS system.39 Presumably, this means that tax return information will be available to Medicaid programs within weeks after returns are filed, just as that information is already available to authorize the payment of other benefits. For example, within two weeks of filing, electronic returns—which are used by 90 percent of EITC claimants40—are available to verify eligibility for Pell Grants and other federally-funded college student aid.41 Along similar lines, IRS sends refund checks within three weeks of taxpayers’ filing of electronic returns.42 Of course, not all low-income consumers file federal income tax returns. To verify eligibility of non-filers, states may be able to access information about prior-year income from quarterly wage records, 1099 tax forms that some employers and banks file with IRS and state tax departments, data from other public agencies showing the receipt of cash assistance, private vendor data, and 8 other sources. But for some consumers who do not file tax returns, state Medicaid programs will be unable to compile a comprehensive picture of prior-year income. The analysis presented below, which shows the levels of prior-year income that verify current eligibility, is thus relevant to many but not all Medicaid applicants and beneficiaries. Current-year income data All else equal, more recent income data—that is, from the current year—can provide a more reliable indicator of current financial circumstances than prior-year IRS filings. States have several options for supplementing prior-year income records with information about earnings during the current year. Under Social Security Act Section 1137, states can access quarterly wage data from their workforce agencies. However, these records are not fully comprehensive: they do not include earnings received as a contractor, earnings from employment in other states, federal earnings, or unearned income. Further, it takes time for state workforce agencies to organize the information and make it available to Medicaid. In California, for example, the Medicaid program cannot obtain data about the first quarter’s wages until the start of the fourth quarter, six months after the first quarter’s end.43 Section 1137 also provides Medicaid programs with reports of new hires within the state; unlike quarterly wage data, these reports are typically available within weeks. However, in addition to the limitations that apply to quarterly wage records, new hires reports are typically unavailable from multi-state employers, who are required to send new hire reports only to a single state.44 Many states obtain near-“real time” wage records for a number of employers, through “The Work Number.” Provided by a subsidiary of the credit reporting firm Equifax, The Work Number makes available information from payroll records of subscribing large employers, which together include more than 2,000 firms. Although this service is costly to states, most Medicaid agencies and other benefit programs used it before the ACA. It will be offered to all states starting in 2014 through the federal data hub. In sum, state Medicaid programs have access to significant but incomplete records about wages received and new employment begun during the current year. Under current law, they will not hear of new hires at multi-state companies, employers across state lines, or federal agencies. Many of these gaps, but not all, could be filled by state or federal contracts with private vendors, like “The Work Number.”45 Initial applications: How Medicaid programs can use past income data to verify attestations of financial eligibility In this section of the paper, we analyze SIPP data to identify the applicants whose attestations of financial eligibility can be verified by data matches that show two things: prior income at Medicaid levels; and a 70 percent or greater likelihood of current monthly income at or below 138 percent FPL, the threshold that will apply in states that expand eligibility. We explore three periods during the calendar year in which Medicaid agencies have access to varying sources of data about consumers’ financial eligibility. For each period, we identify a business rule that meets the above criteria and then assess its potential impact by estimating the percentage of all eligible consumers whose financial eligibility could be verified by the rule. 9 January through April: Verifying financial eligibility when only prior-year income data are available Some Medicaid programs have long permitted the previous year’s income tax return to verify income for applications filed during the first months of the year.46 But prior-year income records will become even more relevant to current Medicaid eligibility in 2014, due to eligibility reaching 138 percent FPL, data-driven verification procedures, and Medicaid’s implementation of MAGI rules that use federal tax definitions to determine financial eligibility for Medicaid. We find that, among consumers whose prior-year incomes did not exceed 138 percent FPL, 85 percent have monthly incomes at Medicaid levels during January through April of the current year (figure 2, table 5). Put differently, if someone’s annual income was low enough to qualify for Medicaid last year, there is an 85 percent chance that the consumer’s monthly income is still below Medicaid’s income eligibility threshold during January through April of the current year. Figure 2. Financial eligibility for Medicaid in January through April among people with prior-year incomes at or below 138 percent FPL Current monthly MAGI of people with prior-year incomes at or below 138 percent FPL Monthly MAGI above Medicaid levels, 15% n=80.5 million Monthly MAGI at Medicaid levels, 85% Source: SIPP 2008. Note: Assumes expanded Medicaid eligibility. In addition, children with current income above 138 percent FPL are counted as financially eligible if they qualify for Medicaid or CHIP under their state’s pre-ACA eligibility standards. Results are limited to U.S. citizens and qualified aliens under age 65. As noted earlier, we are seeking to identify business rules that can verify attestations of Medicaid-level income based on data matches that show prior income at Medicaid levels and that establish at least a 70 percent likelihood of current financial eligibility. These criteria are satisfied by a rule that automatically verifies such attestations in January through April whenever applicants had prior-year income at or below 138 percent FPL, because the probability of having current financial eligibility in that circumstance exceeds 70 percent. Such a rule could have considerable impact. Among all consumers who are eligible for Medicaid based on current monthly income in January through April, fully 79 percent had prior-year income at or below 138 percent FPL (figure 3, table 6). 10 Figure 3. Among all people eligible for Medicaid based on monthly MAGI during January through April, the percentage who had prior-year income at or below 138 percent FPL 100% 90% 80% 70% Percentage of 60% all currently 50% eligible 40% consumers 30% 20% 10% 0% n=86.0 million 79% Prior-year income at or below 138 percent FPL Source: SIPP 2008. Note: See note to figure 2. Not everyone with prior-year income under 138% FPL can verify their prior-year income with federal income tax records, of course. Some indigent consumers do not file returns, as noted earlier. Others may apply for Medicaid before they have filed a federal tax return. Nevertheless, the fact that 79 percent of all eligible consumers during these early months had prior-year incomes low enough for data-based verification suggests that this business rule could greatly streamline enrollment for many if not most eligible consumers who apply during January through April, increasing their participation levels while lowering administrative costs and reducing opportunities for human error. May through August: Verifying financial eligibility when significant new hires data become available During the first six to nine months of the year, depending on the state, Medicaid programs may not have yet gained access to quarterly wage records from the current year. However, states have obtained reports about new hires during the current year, which supplement information about the previous year’s annual income. This combination of data can help Medicaid programs estimate the likelihood of current eligibility. Prior-year income is more likely to predict current income for consumers who did not change jobs since the start of the year. A new job may pay different wages than the old job. This common-sense intuition is confirmed by SIPP data, as shown in table 7. For example, someone with prior-year income at or below 138 percent FPL has an 82 percent likelihood of income at Medicaid levels in May through August of the current year if he or she did not begin a new job before May 1. If he or she did begin such a job, however, the likelihood of current monthly income at Medicaid levels during May through August falls to 67 percent. To meet our criterion and establish a 70 percent or greater likelihood of current monthly eligibility in May through August, data matches would need to show either (a) prior-year income at or below 138 percent FPL and no new job before May 1 of the current year; or 11 (b) prior-year income at or below 100 percent FPL and a new job begun before May 1 of the current year. These two fact combinations would respectively establish an 82 percent and a 70 percent likelihood of monthly income at Medicaid levels during May through August (figures 4 and 5, table 7). Figure 4. Financial eligibility for Medicaid in May through August among people with prior-year incomes at or below 138 percent FPL who did not begin a currentyear new job before May 1 Current monthly MAGI of people with prior-year incomes at or below 138 percent FPL who did not begin a current-year new job before 5/1 Monthly MAGI above Medicaid levels, 18% n = 61.7 million Monthly MAGI at Medicaid levels, 82% Source: SIPP 2008. Note: See note to figure 2 Figure 5. Financial eligibility for Medicaid in May through August among people with prior-year incomes at or below 100 percent FPL who began a current-year new job before May 1 Current monthly MAGI among people with prioryear incomes at or below 100 percent FPL who began a current-year new job before 5/1 Monthly MAGI above Medicaid levels, 30% n = 14.3 million Monthly MAGI at Medicaid levels, 70% Source: SIPP 2008. Note: See note to figure 2 12 In terms of impact, among all consumers who qualify for Medicaid during May through August, 67 percent had prior-year income at Medicaid levels and did not begin a new job before May 1 of the current year. An additional 6 percent had prior-year income at or below 100 percent FPL and began a new job during the current year before May 1 (fig. 6, table 8). Accordingly, a total of 73 percent of eligible consumers could potentially have their attestations of current financial eligibility during May through August verified, using this business rule. Figure 6. Among all people eligible for Medicaid based on monthly MAGI during May through August, the percentage who had various combinations of prior-year income and the start of a current-year new job before May 1 100% 90% 80% 70% 6% 60% Percentage of all currently eligible 50% consumers 40% 30% 67% Prior-year income at or below 100 percent FPL, current-year new job before 5/1 Prior-year income at or below 138 percent FPL, no currentyear new job before 5/1 20% 10% n = 86.0 million 0% Source: SIPP 2008. Note: See note to figure 2. At this juncture, it may be important to remind the reader of an earlier caveat. These results come from survey data gathered during 2009 and 2010. In future years, more people may begin new jobs. If so, the business rule described here would verify attestations of financial eligibility for a smaller proportion of all eligible consumers. September through December: Verifying financial eligibility when wage data become available from earlier quarters It takes time for state workforce agencies to organize their quarterly wage records and make them available in a form that state Medicaid agencies can use. The resulting lag varies among states. In some states, the first quarter’s data is available by the start of the third quarter. In others, including California, it is not available until the fourth quarter. But all states have access to quarterly wage records during the final months of the year. For this analysis, we use earnings during January through April as a proxy for quarterly wage records. We find that, for consumers whose wages during January through April exceeded 80 percent FPL, neither their prior-year income nor whether they began a new job earlier during the current year matters. Consumers with wages that exceeded 80 percent FPL during January 13 through April have less than a 70 percent likelihood of current financial eligibility for Medicaid during September through December, regardless of their other characteristics (table 9). However, among consumers whose wages during January through April did not exceed 80 percent FPL, those with prior-year incomes below 138 percent FPL have an 85 percent likelihood of eligibility (figure 7, table 9). In terms of impact, 55 percent of all consumers who qualify for Medicaid during September through December had prior-year incomes at or below 138 percent FPL and wages during January through April of the current year at or below 80 percent FPL (figure 8, table 10). Figure 7. Financial eligibility for Medicaid in September through December among people with prior-year incomes at or below 138 percent FPL and current-year wages in January through April that did not exceed 80 percent FPL Current monthly MAGI of people with prior-year incomes at or below 138 percent FPL and current-year wages in January through April that did not exceed 80 percent FPL Monthly MAGI above Medicaid levels, 15% Monthly MAGI at Medicaid levels, 85% n = 56.2 million Source: SIPP 2008. Note: See notes to figure 2. Figure 8. Among all people eligible for Medicaid based on monthly MAGI during September through December, the percentage who had prior-year income at or below 138 percent FPL and current-year wages during January through April at or below 80 percent FPL 100% 90% 80% Percentage of 70% 60% all currently 50% eligible 40% consumers 30% 20% 10% 0% n = 86.0 million 55% Prior-year income at or below 138 percent FPL and current-year wages in January-April at or below 80 percent FPL Source: SIPP 2008. Note: See note to figure 2. 14 In using these final results to formulate business rules involving quarterly wage records, state Medicaid agencies need to consider several limitations. For technical reasons involving SIPP data,47 we could not reliably analyze wages received during three-month periods. Instead, we used four-month periods as proxies for calendar quarters. Our goal was to approximate what a state could verify using quarterly wage records from roughly four to seven months in the past. The level of verification shown here could be somewhat higher than what states would achieve with three-month wage records, for several reasons. Four months of wage data may be a more reliable indicator of ongoing income than three months of information can provide. Also, as explained earlier, some quarterly earnings and new hires are outside the records accessible to Medicaid programs. That said, our results are sufficiently powerful to suggest that quarterly wage records, combined with information about prior-year income from all sources, could verify current financial eligibility for a large proportion of applicants who qualify for Medicaid. Redeterminations: How Medicaid programs can use past income data to trigger administrative renewal In this section of the paper, we define the Medicaid beneficiaries whose administrative renewal can be triggered by data matches that show two things: (a) Medicaid-level income during most recent prior period for which information is available; and (b) a 90 percent or greater likelihood of current annual income at Medicaid levels. (As explained earlier, it makes sense to require data to establish a greater likelihood of current financial eligibility for administrative renewal than for verifying initial applications, since beneficiaries experiencing administrative renewal do not make sworn attestations about income like those made by applicants.) As with the previous section, this section identifies, for three periods during the calendar year, business rules that meet the above criteria. We also measure such rules’ potential impact by estimating the percentage of all eligible consumers whose continuing financial eligibility the rules would verify. January through April: Verifying financial eligibility when only prior-year income data are available We find that, among consumers whose prior-year incomes did not exceed 120 percent FPL, 90 percent have Medicaid-level annual incomes during the current year (figure 9, table 11). Consistent with the criteria stated earlier, a state could apply a business rule that administratively renews beneficiaries who had such prior-year income. 15 Figure 9. Annual financial eligibility for Medicaid among people whose prior-year incomes were at or below 120 percent FPL Current annual MAGI above Medicaid levels, 10% Current annual MAGI at Medicaid levels, 90% n = 57.4 million Source: SIPP 2008. Note: See notes to figure 2. This rule could have considerable impact in streamlining redeterminations. Altogether, 71 percent of consumers with current annual income at Medicaid levels had prior-year incomes that did not exceed 120 percent FPL and so could be renewed administratively (figure 10, table 12). Figure 10. Among all people eligible for Medicaid based on current annual MAGI, the percentage that had prior-year income at or below 120 percent FPL 100% 90% 80% 70% Percentage of 60% all currently 50% eligible 40% consumers 30% 20% 10% 0% n = 91.3 million 71% Prior-year income at or below 120 percent FPL Source: SIPP 2008. Note: See note to figure 2. May through August: Verifying financial eligibility when significant new hires data become available Once states can combine prior-year income with reports about new hires during the current year, their ability to predict current annual financial eligibility improves. The likelihood of Medicaid eligibility does not reach 90 percent for people who began a new job during the current year before May 1, regardless of their prior-year income. Among people with prior-year income at or below 120 percent FPL who did not begin such a job, however, 92 percent have current annual income at Medicaid levels (figure 11, table 13). A state could therefore use a business rule that 16 triggers administrative renewal when data matches show that a beneficiary has the latter characteristics. Figure 11. Annual financial eligibility for Medicaid among people whose prior-year incomes were at or below 120 percent FPL and who did not begin a current-year new job before May 1 Current annual MAGI above Medicaid levels, 8% Current annual MAGI at Medicaid levels, 92% n = 62.7 million Source: SIPP 2008. Note: See notes to figure 2. Such a rule could have considerable impact. Altogether, 63 percent of people who qualify for Medicaid based on annual income had prior-year income at or below 120 percent FPL and did not begin a new job during the current year before May 1 (figure 12, table 14). Figure 12. Among all people eligible for Medicaid based on current annual MAGI, the percentage that had prior-year income at or below 120 percent FPL and did not begin a current-year new job before May 1 100% 80% Percentage of 60% all currently eligible 40% consumers 20% 63% 0% n = 91.3 million Prior-year income at or below 120 percent FPL and did not begin a current-year new job before 5/1 Source: SIPP 2008. Note: See note to figure 2. September through December: Verifying financial eligibility when wage data become available from earlier quarters Gaining access to quarterly wage records provides states with very useful information. Much as with initial applications, the likelihood of current annual income at Medicaid levels never exceeds 70 percent for consumers whose wages during January through April of the current year went above 80 percent FPL, regardless of such consumers’ prior-year income. By contrast, almost all consumers (98 percent) have current annual income at Medicaid levels if their wages 17 in January through April were at or below 80 percent FPL and their prior-year income did not exceed 138 percent FPL (figure 13). The same very high likelihood of Medicaid eligibility applies whether or not such consumers began a new job during the current year before September 1 (table 15). Accordingly, a business rule could administratively renew all consumers with prior-year incomes at or below 138 percent FPL and quarterly wages during the current year that did not exceed 80 percent FPL. In terms of impact, such a rule would extend administrative renewal to 60 percent of all consumers who qualify for Medicaid based on annual income (figure 14, table 16). Figure 13. Annual financial eligibility for Medicaid among people whose prior-year incomes were at or below 138 percent FPL and with current-year wages in January through April at or below 80 percent FPL Current annual MAGI above Medicaid levels, 2% Current annual MAGI at Medicaid levels, 98% n = 56.2 million Source: SIPP 2008. Note: See notes to figure 2. Figure 14. Among all people eligible for Medicaid based on current annual MAGI, the percentage whose prior-year income did not exceed 138 percent FPL and whose current-year wages during January through April did not exceed 80 percent FPL 100% 90% 80% Percentage of 70% 60% all currently 50% eligible 40% consumers 30% 20% 10% 0% n = 91.3 million 60% Prior-year income at or below 138 percent FPL and current-year wages at or below 80 percent FPL in January through April Source: SIPP 2008. Note: See note to figure 2. 18 Putting it into practice: How business rules for data-based verification could operate in 2014 The preceding analysis and its underlying calculations are complex. Sorting through SIPP data and fleshing out the relationship between current Medicaid eligibility, defined in terms of monthly and annual income, and prior-year income, the start of new employment during the current year, and quarterly wages earned at earlier periods of the current year requires careful attention to detail and can be tedious. The result of this hard work, however, can be a straightforward set of business rules specifying the data matches that eliminate the need for traditional verification. States already use such business rules in many contexts, but they should become an increasingly prominent feature of Medicaid under the ACA’s data-driven eligibility approach. Each state’s rules will reflect its policy choices about budget periods, probability thresholds, and other matters. To simplify our presentation, the above analysis sought to develop rules for a state that makes the following choices: Verification rules involving past financial data and current financial eligibility for MAGIbased coverage will not treat adults differently than children; Financial eligibility for initial applications will be based on current monthly income; Financial eligibility for redeterminations will be based on projected annual income during the current year; When an applicant attests to current monthly income at Medicaid levels, “reasonable compatibility” is defined so the attestation can be verified by data that: (a) shows prior income at Medicaid levels; and (b) establishes at least a 70 percent likelihood of current income at Medicaid levels; and When a beneficiary is slated for redetermination, administrative renewal will be triggered by data that: (a) shows prior income at Medicaid levels; and (b) establishes at least a 90 percent likelihood of current income at Medicaid levels. In 2014, such a state could use the business rules shown in Tables 1 and 2, which summarize the results of our analysis. 19 Table 1. Possible business rules for initial applications in 2014: applicants whose attestations of current Medicaid-level income are verified by data matches Time frame Data that verify an attestation of current financial eligibility January through April 2014 May through August 2014 2013 income no higher than 138 percent FPL Either: (a) 2013 income no higher than 138 percent FPL and no new job between 1/1/2014 and 5/1/2014; or (b) 2013 income no higher than 100 percent FPL and a new job begun between 1/1/2014 and 5/1/2014 Sept. 2013 income at or below 138 percent through FPL and quarterly wages, earlier Dec. 2014 during 2014, at or below 80 percent FPL Based on that data, the likelihood of current monthly income at Medicaid levels 85% (a) 82% Among all Medicaideligible consumers, the percentage potentially verified by the rule 79% 73% (b) 70% 85% 55% Source: SIPP 2008. Note: Business rules assume financial eligibility based on current monthly income and verification based on data matches that (a) show prior income at or below 138 percent FPL and (b) establish a 70 percent or greater likelihood of current income at or below 138 percent FPL. Table 2. Possible business rules for renewal in 2014: beneficiaries whose eligibility is administratively renewed Time frame Data match that triggers administrative renewal January through April 2014 May through August 2014 Sept. through Dec. 2014 2013 income no higher than 120 percent FPL Based on that data match, the likelihood of 2014 annual income at Medicaid levels 90% 2013 income no higher than 92% 120 percent FPL and no new job between 1/1/2014 and 5/1/2014 2013 income at or below 138 98% percent FPL and quarterly wages, earlier during 2014, at or below 80 percent FPL Among all Medicaideligible consumers, the percentage potentially renewed administrative by the rule 71% 63% 60% Source: SIPP 2008. Note: Business rules assume financial eligibility based on current annual income and verification based on data matches that (a) show prior income at or below 138 percent FPL; and (b) establish a 90 percent or greater likelihood of current income at or below 138 percent FPL. 20 Of course, states could make other choices. To illustrate how states could use our findings to develop business rules based on such choices, we develop an alternate set of rules, premised on various state decisions about Medicaid eligibility, some of which we keep the same as before: Rules that use prior financial data to verify MAGI-based eligibility will not treat adults differently than children; and Financial eligibility for initial applications will be based on current monthly income. However, we modify other state choices: Financial eligibility for redeterminations will be based on current monthly income at the time of redetermination, rather than on projected current annual income; When an initial applicant attests to Medicaid-level income, reasonable compatibility is defined so the attestation can be verified by data that (a) show prior income at Medicaid levels and (b) establish at least a 60 percent likelihood of current income at Medicaid levels, rather than a 70 percent likelihood; and When a beneficiary is slated for redetermination, administrative renewal will be triggered by data that (a) show prior income at Medicaid levels and (b) establish more than an 85 percent likelihood of financial eligibility, rather than a likelihood of 90 percent or more.48 Tables 3 and 4 show the alternate business rules that result from these state choices. Table 3. Alternate 2014 business rules for verifying applicant attestations of Medicaid-level income Based on that data, the likelihood of current monthly income at Medicaid levels1 Impact: Among all Medicaid-eligible consumers, the percentage verified by the rule2 Time frame Data that verifies an attestation of current financial eligibility January through April 2014 May through August 2014 2013 income no higher than 138 percent FPL 85% 79% 2013 income no higher than 138 percent FPL 75% Sept. through Dec. 2014 2013 income at or below 138 percent FPL plus either (a) quarterly wages, earlier during 2014, at or below 80 percent FPL; or (b) no new job between 1/1/2014 and 9/1/2014 82% for people not beginning a new job during the current year before 5/1; 67% for people beginning such a job (a) 85% 67% (b) 60% Source: SIPP 2008. Note: Business rules assume financial eligibility based on current monthly income and verification based on data matches that (a) show prior income at or below 138 percent FPL and (b) establish a 60 percent or greater likelihood of current income at or below 138 percent FPL. 1. See tables 5, 7, and 9 for these estimates of the likelihood of current eligibility. 2. See tables 6, 8, and 10 for these estimates of impact. 21 Table 4. Alternate 2014 business rules for administrative renewal Based on that data match, the likelihood of current monthly income at Medicaid levels1 Impact: Among all Medicaid-eligible consumers, the percentage who are administratively renewed by the rule2 Time frame Data match that triggers administrative renewal January through April 2014 May through August 2014 Sept. through Dec. 2014 2013 income no higher than 120 percent FPL 87% 73% 2013 income no higher than 100 percent FPL and no new job between 1/1/2014 and 5/1/2014 87% 55% 2013 income at or below 138 percent FPL, and quarterly wages, earlier during 2014, at or below 80 percent FPL, and no new job between 1/1/2014 and 9/1/2014 88% 45% Source: SIPP 2008. Note: Business rules assume financial eligibility based on current monthly income at the time of redetermination and verification based on data matches that (a) show prior income at or below 138 percent FPL and (b) establish more than an 85 percent likelihood of current income at or below 138 percent FPL. 1. See tables 5, 7, and 9 for these estimates of the likelihood of current eligibility. 2. See tables 6, 8, and 10 for these estimates of impact. The alternate rules for initial verification yield results that differ from the original business rules in unsurprising ways. The alternate rules change the reasonable compatibility definition so that a data match establishing a 60 percent likelihood of current eligibility (instead of a 70 percent likelihood, as in the original rules) can verify an attestation of financial eligibility. As a result, more people can have their attestations verified by data matches. During January through April, the proportion of eligible consumers who can be verified is unchanged, at 79 percent; in May through August, it rises slightly, from 73 percent under the original rules to 75 percent under the alternate rules; and during September through December, the increase is more pronounced, from 55 percent to 67 percent (tables 1 and 3). By contrast, the results for administrative renewal are somewhat unexpected. One factor makes administrative renewal more likely: the alternate rules lower the verification threshold so that, instead of requiring at least a 90 percent likelihood of continuing eligibility, data must show more than an 85 percent likelihood of current eligibility. But moving from annual to monthly budget periods makes such renewal less likely. It turns out that the second change is far more consequential. During January through April, slightly more eligible consumers can be renewed administratively under the alternate rules—73 percent, rather than 71 percent as under the original rules (tables 2 and 4). However, for May through August, the alternate rules allow administrative renewal for 55 percent of eligible consumers, rather than 66 percent under the original rules; and in September through December, the percentage that can be administratively renewed drops from 60 percent under the original rules to 45 percent (tables 2 and 4). 22 These are relative subtleties, however. The most important point is that, once a state makes basic decisions about its approach to Medicaid eligibility, it can use our SIPP findings to craft business rules that clearly and straightforwardly identify the circumstances under which eligibility can be verified based on data matches, without any need for consumer documentation or caseworker involvement. Conclusion Early in the calendar year, prior-year income data are “fresh” and most directly relevant in verifying current financial eligibility. As time goes by, those data age and become less informative about households’ present circumstances; however, more recent information becomes available about events during the current year. We find that, in every month, the combination of prior-year income data and current-year new hires data or quarterly wage reports furnishes enough information to verify eligibility for more than half of consumers who qualify for Medicaid. In such cases, establishing eligibility based on data matches rather than documentation from consumers could save administrative resources, lessen burdens on lowincome households, increase participation levels among eligible consumers, and reduce the proportion of mistaken eligibility outcomes. To be sure, many consumers will still require old-fashioned, manual verification of eligibility. And the transition to more data-driven methods will not be instantaneous. But policymakers have good reason to hope that, between the income and employment records described here and other data sources that are outside the scope of this paper or that will become available as the country’s information technology revolution continues to unfold, a substantially more humane, efficient, and accurate Medicaid eligibility system lies ahead. 23 ADDITIONAL TABLES 24 Table 5. Among people under age 65 with various levels of prior-year income, the percentage who are financially eligible for Medicaid, based on current monthly income during January through April Prior-year income 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL Percentage of consumers who qualify based on current monthly MAGI during January through April 95% 91% 89% 87% 85% How to read this figure: Among people under age 65 who received no income during the prior year, 95% have monthly income at Medicaid levels (i.e., at or below 138 percent FPL) during January through April of the current year; among people who received income between 0 and 80 FPL last year, 91 percent have Medicaid-level monthly income during January through April of the current year; etc. Source: 2008 SIPP. Note: Income is MAGI. If MAGI is at or below 138 percent FPL, income is at Medicaid levels, since this analysis assumes that states implement the ACA Medicaid expansion. In addition, children with current income above 138 percent FPL are counted as eligible if they qualify for Medicaid or CHIP under their state’s pre-ACA eligibility standards. Results are limited to U.S. citizens and qualified aliens. Table 6. Among all people under age 65 who qualify for Medicaid based on monthly MAGI during January through April, the percentage with prior-year income at various levels Prior-year income 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL Among consumers who qualify based on current monthly MAGI during January through April, the percentage who had prior-year income at specified levels 12% 55% 64% 73% 79% How to read this table: Among all consumers who qualify for Medicaid based on monthly MAGI in January through April, 12% had prior-year income at 0 percent FPL; 55% had prior-year income at 0-80 percent FPL; 64% had prior-year income at 0-100 percent FPL; etc. Source: 2008 SIPP. Note: See notes to table 5. 25 Table 7. Among people under age 65 with various levels of prior-year income and certain other characteristics, the percentage who are financially eligible for Medicaid, based on current monthly income during May through August Prior-year income and new job during current year 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL Eligible based on current monthly MAGI during May through August New job before 5/1 No new job before 5/1 New job before 5/1 No new job before 5/1 New job before 5/1 No new job before 5/1 New job before 5/1 No new job before 5/1 New job before 5/1 No new job before 5/1 * 92% 72% 88% 70% 87% 68% 85% 67% 82% How to read this Table: Among people who received no income during the prior year and, during the current year, did not begin a new job before 5/1, 92 percent have Medicaid-level monthly incomes in May through August; among people who received income between 0 and 80 percent FPL during the prior year and, in the current year, began a new job before 5/1, 72 percent have Medicaid-level incomes during May through August; etc. Source: 2008 SIPP. Note: See notes to table 5. Cells with * have too few people in the sample to yield meaningful results. Table 8. Among all consumers under age 65 who qualify for Medicaid based on monthly MAGI during May through August, the percentage with prior-year income at various levels and other characteristics Among consumers who qualify based on monthly MAGI in May through August Prior-year income Prior-year income at specified levels 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL 12% Prior-year income at specified levels, plus no current-year new hire before 5/1 11% Prior-year income at specified levels, plus a current-year new hire before 5/1 1% 53% 47% 6% 61% 55% 6% 69% 62% 7% 75% 67% 8% How to read this table: Among all consumers who qualify for Medicaid based on monthly MAGI in May through August: (a) 12% had prior-year income at 0 percent FPL, including (i) 11% who did not begin a new job during the current year before May 1 and (ii) 1 percent who did begin such a job; (b) 53 percent had prioryear income at 0-80 percent FPL, including (i) 47 percent who did not begin a new job before May 1 of the current year and (ii) 6 percent who did begin such a job; etc. Source: 2008 SIPP. Note: See notes to table 5. 26 Table 9. Among people under age 65 with various levels of prior-year income and certain other characteristics, the percentage who are financially eligible for Medicaid, based on current monthly MAGI during September through December Earnings during January through April of the current year Prior-year income and the start of new employment during the current year 0 percent FPL 0-80 percent FPL At or below 80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL 0 percent FPL 0-80 percent FPL More than 80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 Eligible based on current monthly MAGI during September through December 77% 95% 75% 89% 75% 89% 73% 89% 73% 88% Combined: 93% Combined: 86% Combined: 86% Combined: 85% Combined: 85% * * 42% 63% 47% 64% 48% 64% 49% 60% How to read this table: Among people whose wages during January through April of the current year were at or below 80 percent FPL, who received no income during the prior year, and who began a new job before 9/1 of the current year, 77 percent have Medicaid-level monthly incomes during September through December of the current year; among people whose wages during January through April of the current year were at or below 80 percent FPL, who received no income during the prior year, and who not begin a new job before 9/1 of the current year, 95 percent have Medicaidlevel monthly incomes during September through December of the current year; and among all people whose wages during January through April of the current year were at or below 80 percent FPL and who received no income during the prior year, including both those who did and those who did not begin a new job before 9/1 of the current year, 93 percent have Medicaid-level monthly incomes during September through December of the current year; etc. Source: 2008 SIPP. Note: See notes to table 5. Cells with * have too few people in the sample to yield meaningful results. 27 Table 10. Among all consumers under age 65 who qualify for Medicaid based on monthly MAGI during September through December, the percentage with prioryear income at various levels and other characteristics Among consumers who qualify based on current monthly MAGI during September through December Prioryear income Current-year wages during Jan. through April at or below 80 percent FPL 11% 10% 1% Did not begin a new job before 9/1 of the current year Current-year wages during Jan. All through April above 80 percent FPL 10% * 34% 27% 7% 31% 4% 1% 50% 41% 9% 47% 6% 2% 53% 43% 10% 53% 10% 3% 55% 45% 10% 57% 12% 4% All 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL No currentyear new job before 9/1 Currentyear new job before 9/1 Current-year wages during Jan. through April above 80 percent FPL and began current-year new job before 9/1 * How to read this table: The first row shows that, among all consumers who qualify for Medicaid based on monthly MAGI in September through December: (a) 11% had prior-year income at 0 percent FPL and current-year wages during January through April not above 80 percent FPL, including 10% who did not begin a new job before September 1 of the current year and 1% who did begin such a job; and (b) 10% had prior-year income at 0 percent FPL and did not begin a new job before September 1 of the current year (including those with wages above and those with wages below 80 percent FPL during January through April). The second row shows that, among all consumers who qualify for Medicaid based on monthly MAGI in September through December: (a) 34% had prior-year income between 0 and 80 percent FPL and current-year wages during January through April not above 80 percent FPL, including 27% who did not begin a new job during the current year before September 1 and 7% who did begin such a job; (b) 31% had prior-year income between 0 and 80 percent FPL and began a job before September 1 of the current year, including 4% whose current-year wages during January through April exceeded 80 percent FPL; and (c) 1 percent had prior-year income between 0 and 80 percent FPL, currentyear wages during January through April above 80 percent FPL, and began a new job during the current year before 9/1. Source: 2008 SIPP. Note: See notes to table 5. 28 Table 11. Among people under age 65 with various levels of prior-year income, the percentage who are financially eligible for Medicaid, based on current annual income Prior-year income Percentage of consumers who qualify based on current annual MAGI 96% 93% 92% 90% 88% 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL How to read this figure: Among people under age 65 who received no income during the prior year, 96% have current annual income at Medicaid levels (i.e., at or below 138 percent FPL); among people who received income between 0 and 80 FPL last year, 93 percent have Medicaidlevel annual income during the current year; etc. Source: 2008 SIPP. Note: See notes to table 5. Table 12. Among people under age 65 who qualify for Medicaid based on annual income, the percentage with prior-year income at various levels Prior-year income 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL Among all consumers who qualify based on current annual MAGI, the percentage who had prior-year income at specified levels 11% 54% 63% 71% 78% How to read this table: Among all consumers who qualify for Medicaid based on current annual income, 11% had prior-year income at 0 percent FPL; 54% had prior-year income at 0-80 percent FPL; 63% had prior-year income at 0-100 percent FPL; etc. Source: 2008 SIPP. Note: See notes to table 5. 29 Table 13. Among people under age 65 with various levels of prior-year income and certain other characteristics, the percentage who are financially eligible for Medicaid, based on current annual income Prior-year income and new job during current year Eligible based on current annual MAGI 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL New job before 5/1 No new job before 5/1 New job before 5/1 No new job before 5/1 New job before 5/1 No new job before 5/1 New job before 5/1 No new job before 5/1 New job before 5/1 No new job before 5/1 * 96% 85% 94% 83% 93% 82% 92% 80% 89% How to read this Table: Among people who received no income during the prior year and, during the current year, did not begin a new job before 5/1, 96 percent have Medicaid-level annual incomes during the current year; among people who received income between 0 and 80 percent FPL during the prior year and, during the current year, began a new job before 5/1, 85 percent have Medicaid-level annual incomes during the current year; etc. Source: 2008 SIPP. Note: See notes to table 5. Cells with * have too few people in the sample to yield meaningful results. Table 14. Among all consumers under age 65 who qualify for Medicaid based on current annual income, the percentage with prior-year income at various levels and other characteristics Among consumers who qualify based on current annual MAGI Prior-year income Prior-year income at specified levels Prior-year income at specified levels, plus no current-year new hire before 5/1 Prior-year income at specified levels, plus a current-year new hire before 5/1 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL 11% 11% 0% 54% 48% 6% 63% 56% 7% 71% 63% 8% 78% 68% 10% How to read this table: Among all consumers who qualify for Medicaid based on current year MAGI: 11% had prior-year income at 0 percent FPL, none of whom began a new job during the current year; 54% had prior-year income at 0-80 percent FPL, including 48% who did not begin a new job during the current year before May 1 and 6% who did begin such a job; etc. Source: 2008 SIPP. Note: See notes to table 5. 30 Table 15. Among people under age 65 with various levels of prior-year income and certain other characteristics, the percentage who are financially eligible for Medicaid, based on current annual income Earnings during January through April of the current year Prior-year income and the start of new employment during the current year 0 percent FPL 0-80 percent FPL At or below 80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL 0 percent FPL 0-80 percent FPL More than 80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 New job before 9/1 No new job before 9/1 Eligible based on current annual MAGI 99% 99% 99% 99% 99% 99% 98% 98% 98% 98% * * 57% 66% 59% 69% 59% 70% 59% 68% How to read this table: Using the final two rows as an example, among people whose wages during January through April of the current year exceeded 80 percent FPL, who had prior-year income between 0 and 138 percent FPL, and who began a new job during the current year before 9/1, 59 percent have current annual incomes at Medicaid levels; among people whose wages during January through April of the current year exceeded 80 percent FPL, who had prior-year income between 0 and 138 percent FPL, but who did not begin a new job during the current year before 9/1, 68 percent have current annual incomes at Medicaid levels. Source: 2008 SIPP. Note: See notes to table 5. Cells with * have too few people in the sample to yield meaningful results. 31 Table 16. Among all consumers under age 65 who qualify for Medicaid based on current annual income, the percentage with prior-year income at various levels and other characteristics Among those who qualify based on current annual MAGI Prioryear income 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL Current-year wages during Jan. through April at or below 80 percent FPL Did not begin a new job before 9/1 of the current year Current-year wages during Jan. All through April above 80 percent FPL Current-year wages during Jan. through April above 80 percent FPL and current-year new job before 9/1 All No current year new job before 9/1 Current year new job before 9/1 11% 10% 1% 10% * * 37% 28% 9% 32% 4% 2% 54% 42% 12% 49% 7% 2% 58% 45% 13% 56% 11% 3% 60% 47% 13% 60% 13% 4% How to read this table: To use the final row as an example, among all consumers who qualify for Medicaid based on annual MAGI: (a) 60% had prior-year income between 0 and 138 percent FPL and current-year wages during January through April not above 80 percent FPL, including 47% who did not begin a new job before September 1 of the current year and 13% who did begin such a job; (b) 60% had prior-year income between 0 and 138 percent FPL and did not begin a new job in the current year before 9/1, including 13 percent whose wages during January through April of the current year exceeded 80 percent FPL; and (c) 4 percent had prior-year income between 0 and 138 percent FPL, wages during January through April of the current year that exceeded 80 percent FPL, and began a new job in the current year before 9/1. Source: 2008 SIPP. Note: See notes to table 5. Cells with * have too few people in the sample to yield meaningful results. 32 Appendix A: California findings We took two different approaches to replicating our analysis for residents of California, the nation’s most populous state. First, we took the national SIPP sample and “reweighted” it to reflect the characteristics of Californians, rather than Americans as a whole. The results of this analysis were virtually indistinguishable from our results for all U.S. residents. Second, we analyzed the SIPP respondents who lived California. We found some differences with the remainder of the SIPP sample, particularly with respect to the earnings of part-time workers. However, we were unable to determine whether such divergence reflected actual population differences or the chance characteristics of the Californians who happened to be in the sample. Highlighting the uncertainty created by a reduced sample size, Californians in the SIPP with prior-year income between 101-120 percent FPL were more likely to have income at Medicaid levels during September through December of the current year than during May through August. The same was true for Californians with incomes between 121-138 percent FPL. These are implausible results that almost certainly reflect the chance features of the California SIPP sample rather than real characteristics of Californians as a whole. The key question facing California policymakers is whether our SIPP results are a reasonable guide to the relationship between past and current financial eligibility for Medi-Cal in California, under the ACA’s 2014 rules. Differences between California respondents and other SIPP respondents are most likely to reflect real population characteristics, rather than the chance features of the SIPP sample, if those differences emerge from significant aggregations of data— that is, if they are seen in comparisons between large groups of survey respondents, each of whom provided multiple survey responses. We therefore examined whether California respondents differed from SIPP respondents nationally in terms of the relationship between prior-year income and current annual income, without factoring in new hires and partial-year wage data. This comparison between large aggregations of survey reports yielded results shown in Table A1. The modest differences suggest that verification rules based on national SIPP data should be applicable in California. Table A1. Among consumers with prior-year incomes at various levels, the percentage with current annual income at Medicaid levels: U.S. vs. California Prior-year income U.S. California 0 percent FPL 0-80 percent FPL 0-100 percent FPL 0-120 percent FPL 0-138 percent FPL 96% 93% 92% 90% 88% 92% 94% 92% 90% 87% How to read this table: Among U.S. and California residents with prior-year incomes at 0 percent FPL, 96 percent and 92 percent, respectively, have current-year income at Medicaid levels; etc. Source: 2008 SIPP. Note: This table contrasts the characteristics of all SIPP respondents nationally with those of SIPP respondents living in California. As noted, a comparison between U.S. residents and Californians that reweighted the national SIPP sample did not find significant differences. 33 Appendix B. Methodology These results are based on the 2008 Survey of Income and Program Participation (SIPP). The SIPP collects data in waves from a panel of the same households over several years. Each SIPP wave lasts four months. Households are interviewed once per wave, when they provide information about the current month as well as the preceding three months. Households enter the survey on a staggered basis, so that approximately one-fourth of the households in the panel are interviewed in each month. Our data include all 10 waves of the 2008 SIPP. The SIPP groups individuals into households, and within households, families, subfamilies, and individuals. Because none of these groupings reflects the set of people whose income and other attributes are aggregated for the purposes of determining Medicaid eligibility under the ACA, we constructed an alternative grouping, the Medicaid unit (MDCDU) for these purposes. There were two steps to creating Medicaid units. First, we broke households into standard health insurance units as best we could. Second, we rearranged these health insurance units into Medicaid units. To construct health insurance units (HIUs), we first looked to units consisting of a single individual. In cases where an individual is not a member of a family, we classified that individual as her own health insurance unit. This case occurred automatically if the individual was alone in a household, but also occurred, for example, if an individual appeared by herself in a household that also contained a separate family. Next, the members of a family or subfamily were also considered a health insurance unit. The longitudinal key of the individual or the longitudinal key of the reference person of a family was used as a persistent indicator of a particular health insurance unit over time. To assign observations to Medicaid units, we first classified each individual in each wave by tax filing status, using the exhaustive income data available within the SIPP. We then catalogued all health insurance units based on whether they consisted of at least one filer. Then, in households where at least one HIU contained a filer and at last one HIU did not, we decomposed HIUs with zero filers and assigned each member to an HIU with at least one filer, if we could identify a relationship between that member and one of the members of an HIU with at least one filer. In most cases, there was only one possible match. In cases of multiple HIUs with filers within the household, we prioritized, in hierarchical order, relationships with (1) individuals who were the reference person over individuals who were not, (2) filers over those who were not, (3) spouses over parents and children, and (4) the old over the young. Medicaid eligibility under the ACA will be based on modified adjusted gross income (MAGI) as defined in the law. The most important components of MAGI are all available from SIPP survey responses. Adults were considered eligible for Medicaid if MAGI for the relevant period was at or below 138 percent of the FPL. We did not model state-specific programs that currently extend eligibility for nonelderly adults above this level. Maintenance of eligibility for adults above 138 percent of the FPL is not required beginning in 2014, so many of these programs are likely to cease. For children, we modeled eligibility under both the Medicaid expansion threshold of 138 percent of the FPL and the higher pre-ACA thresholds in each state, which include CHIP. We produced aggregated characteristics for each MDCDU for use in our data mining efforts. 2009 aggregate wages, age of the oldest member of the MDCDU in 2009, indicator variables for the receipt of veteran’s benefits, child support, SSI, public assistance, and workers’ compensation in 2009 within the household were produced based on survey responses. 2009 MAGI was based on starting MDCDU size, the 2009 poverty line thresholds, and aggregated 34 MDCDU MAGI constructed in the standard way. Medicaid units were classified into three types, with these determinations made during the first wave only: (1) those with only one person (“single, with no dependents”), (2) those containing entirely filers, and (3) those containing some individuals not classified as filers. Any worker and any full-time worker were based on the presence of individuals with those employment statuses in 2009. To find the groups of adults and children most likely to retain eligibility (as shown in Tables C5 and C6, in Appendix C), we used data mining techniques. Data mining is a collection of algorithms and techniques from computer science and statistics that are used to discover patterns in data. We analyzed data consisting of characteristics such as those described above plus a classifying variable indicating whether or not the person would be eligible for Medicaid based on 2010 annual income (or income for a monthly reporting period in 2010) for those who started off being eligible in 2009. We then apply a classifying algorithm, which identifies combinations of characteristics associated with high rates of 2010 eligibility. We used Orange, a free, opensource machine learning software package written in the Python programming language.49 Several algorithms were tested. The groups that we report were identified by the classification tree algorithm. Since our goal was to predict eligibility in 2010 using data from 2009 and early 2010, it was necessary to limit our analysis to observations for which there were data available over two years. There is attrition over time in the SIPP, which excluded some observations were excluded. We reweighted the remainder so that they would still be representative of both California and the nation in terms of important demographic and economic variables such as income, age, gender, and race/ethnicity. Specifically, we reweighted to match the distribution of these characteristics in American Community Survey, which has a far larger sample size than the SIPP, to replicate the most accurate data publicly available about the low-income population. An important limitation in the SIPP data is “seam bias.” Questions are asked only once every four months, resulting in errors in responses for months other than the ones during which questions were asked.50 Ideally, we would have produced results for particular months or for quarters during the year, rather than four-month periods. However, we found that using months other than the interview months did, in fact, distort the resulting patterns of eligibility. To have large enough samples for meaningful results, we needed to group Medicaid eligibility months into the four-month segments that make up each SIPP wave, establishing current monthly income levels based exclusively on the months during which questions were asked. Also, since our data were for the period from 2008 to 2010, they reflect underlying economic trends during that period. What we term “prior-year” income is income during calendar year 2009, as shown by the 2008 SIPP. What we call “current” income was realized during calendar year 2010. If economic trends after January 2014 involve a notable increase in employment and/or wages for the poor and lower middle-class, the trends in eligibility over time could differ from the results presented here. 35 Appendix C. Additional results Table C1. The likelihood of current income at Medicaid levels, by prior-year, overlapping income levels, for people under age 65 Group defined by prior-year income 0 percent FPL 0 to 80 percent FPL 0 to 100 percent FPL 0 to 120 percent FPL 0 to 138 percent FPL Number Likelihood of current monthly income at of people or below 138 percent FPL group January May September (millions) through through through April August December Likelihood of current annual income at or below 138 percent FPL 11.0 52.5 62.3 95% 91% 89% 91% 86% 85% 90% 82% 81% 96% 93% 92% 72.0 87% 83% 79% 90% 80.5 85% 80% 76% 88% Source: 2008 SIPP. Note: Income is MAGI. Results are limited to U.S. citizens and qualified aliens. Table C2. The likelihood of current income at Medicaid levels, by prior-year, non-overlapping income levels, for people under age 65 Group defined by prior-year income 0 percent FPL 1 to 80 percent FPL 81 to 100 percent FPL 101 to 120 percent FPL 121 to 138 percent FPL Number Likelihood of current monthly income at of people or below 138 percent FPL group January May September (millions) through through through April August December Likelihood of current annual income at or below 138 percent FPL 11.0 41.6 95% 90% 91% 85% 90% 80% 96% 93% 9.8 80% 77% 71% 85% 9.8 76% 70% 67% 80% 8.4 64% 60% 56% 67% Source: 2008 SIPP. Note: Income is MAGI. Results are limited to U.S. citizens and qualified aliens. 36 Table C3. The likelihood of current income at Medicaid levels during May through August and annually, by prior-year, non-overlapping income and the start of a new job before May 1 of the current year, for people under age 65 Group defined by prior-year income and current-year new hire before May 1 0 percent FPL New hire No new hire 1 to 80 percent FPL New hire No new hire 81 to 100 percent FPL New hire No new hire 101 to 120 percent FPL New hire No new hire 121 to 138 percent FPL New hire No new hire Number of people in group (millions) Likelihood of monthly income at Medicaid levels during May through August Likelihood of annual income at Medicaid levels 0.6 10.4 * 92% * 96% 5.9 35.6 73% 87% 85% 94% 1.4 8.3 63% 79% 76% 87% 1.4 8.4 55% 73% 75% 81% 1.1 7.3 56% 60% 62% 68% Source: 2008 SIPP. Note: Income is MAGI. Results are limited to U.S. citizens and qualified aliens. An asterisk indicates a sample size too small for reliable results. Table C4. The likelihood of current income at Medicaid levels during September through December and annually, by prior-year, non-overlapping income and the start of a new job before September 1 of the current year, for people under age 65 Group defined by prioryear income and currentyear new hire before September 1 0 percent FPL New hire No new hire 1 to 80 percent FPL New hire No new hire 81 to 100 percent FPL New hire No new hire 101 to 120 percent FPL New hire No new hire 121 to 138 percent FPL New hire No new hire Number of people in group (millions) Likelihood of monthly income at Likelihood of annual Medicaid levels during income at Medicaid September through December levels 1.3 9.6 72% 93% 91% 96% 10.4 31.1 68% 85% 89% 94% 2.5 7.3 62% 74% 80% 87% 2.3 7.4 56% 70% 78% 81% 2.2 6.2 61% 55% 69% 66% Source: 2008 SIPP. Note: Income is MAGI. Results are limited to U.S. citizens and qualified aliens. 37 Table C5. Various groups of U.S. children whose prior-year characteristics establish a high likelihood of current income at Medicaid levels Likelihood of current monthly income at or below 138 percent FPL Likelihood of current annual income at or below 138 percent FPL Income and other characteristics during the prior year Number in the prior-year January group (millions) through April May through August September through December 0-30% FPL, no full-time worker 0-84% FPL, received TANF 0-84% FPL, no TANF 31-118% FPL, no fulltime worker 85-118% FPL 6.9 95% 93% 94% 99% 2.9 91% 92% 90% 99% 16.4 5.9 89% 87% 88% 87% 86% 82% 95% 95% 5.9 79% 78% 73% 87% How to read this table: During the prior year, 6.9 million children lived in families who had annual incomes at or below 30 percent FPL, without any full-time workers at any point in the year. Among those children, the proportions with current monthly income at or below 138 percent FPL were 95 percent during January through April, 93 percent during May through August, and 94 percent in September through December. Within that same group of children, 99 percent had current annual income at or below 138 percent FPL. Source: 2008 SIPP. Note: If MAGI is at or below 138 percent FPL, income is at Medicaid levels. This table understates the likelihood of current eligibility since it does not include some current eligibility for children in states that provided Medicaid or CHIP coverage above 138 percent FPL before ACA; such coverage must remain in effect until 2019, under the ACA’s maintenance of effort requirements. Results are limited to U.S. citizens and qualified aliens. A child is listed in the “full-time worker” group if a parent or guardian in the household worked full-time at some point during the previous year. This table lists children as TANF recipients if they received public assistance in the form of TANF cash aid or general assistance at some point during the prior year. 38 Table C6. Various adults whose prior-year characteristics establish a high likelihood of current income at Medicaid levels Income and other characteristics during the prior year 0-27% FPL, no full-time worker 28-84% FPL, receives SSI 0-23% FPL, parent, fulltime worker 28-70% FPL, no SSI 28-84% FPL, parent, fulltime worker, 85-120% FPL, no workers Number in the prioryear group (millions) Likelihood of current monthly income at or below 138 percent FPL Likelihood of current annual income at or below 138 percent FPL January through April May through August September through December 16.9 95% 91% 89% 97% 6.6 1.3 92% 87% 83% 79% 77% 75% 93% 90% 0.4 14.1 90% 76% 97% 71% 90% 67% 96% 80% 1.8 81% 68% 69% 81% How to read this table: During the prior year, 16.9 million adults received annual income at or below 27 percent FPL in a household that included no full-time workers for any part of the year. Among those adults, the proportions with current monthly income at or below 138 percent FPL are 95 percent during January through April, 91 percent during May through August, and 89 percent in September through December. Within that same group of adults, 97 percent have current annual income at or below 138 percent FPL. Source: 2008 SIPP. Note: All listed groups have at least a 75 percent likelihood of current annual income at Medicaid levels. If MAGI is at or below 138 percent FPL, income is at Medicaid levels, as this analysis assumes that states implement the ACA Medicaid expansion. Results are limited to U.S. citizens and qualified aliens. An adult falls within the “full-time worker” group if either the adult or the adult’s spouse worked full-time at some point during the previous year. An adult falls within the “part-time worker” group if either the adult or the adult’s spouse worked part-time at some point during the previous year and neither worked full-time at any point during the previous year. This table lists people as TANF recipients if they received public assistance in the form of TANF cash aid or general assistance at some point during the prior year. About the Authors and Acknowledgments Stan Dorn and Matthew Buettgens are a senior fellow and a senior research associate, respectively, at the Urban Institute’s Health Policy Center. Habib Moody and Christopher Hildebrand are research assistants. The authors are grateful to the California HealthCare Foundation for supporting our research. In addition, we greatly appreciate the thoughtful comments about earlier drafts of this report provided by Catherine Teare of the Foundation, comments from Sophie Snyder of The Children’s Partnership on earlier drafts of the report, as well as insights from Lisa Dubay of the Urban Institute about how best to analyze the SIPP data and her review of earlier drafts of this report. Finally, the authors would like to thank Dr. Len Finocchio, formerly Associate Director of the California Department of Health Care Services and Senior Program Officer for the California HealthCare Foundation, who suggested the idea for our data match project, including this report. Neither those individuals, the Foundation, nor the Urban Institute, its funders, or trustees are responsible for the opinions expressed in this report, which are the authors’. 39 About the Urban Institute The Urban Institute gathers data, conducts research, evaluates programs, offers technical assistance overseas, and educates Americans on social and economic issues—to foster sound public policy and effective government. We build knowledge about the nation’s social and fiscal challenges, practicing open-minded, evidence-based research to diagnose problems and figure out which policies and programs work best, for whom, and how. About the California HealthCare Foundation The California HealthCare Foundation, based in Oakland, California, works as a catalyst to fulfill the promise of better health care for all Californians. We support ideas and innovations that improve quality, increase efficiency, and lower the costs of care. Notes 1 In determining eligibility, 5 FPL percentage points are subtracted from modified adjusted gross income (MAGI). If the resulting net income level does not exceed 133 percent FPL, the consumer is financially eligible for Medicaid. As a result, the gross income threshold for Medicaid eligibility is 138 percent FPL. 2 National Federation of Independent Business v. Sebelius, 132 S. Ct. 2566 (2012). 3 Kaiser Commission on Medicaid and the Uninsured. “Status of State Action on the Medicaid Expansion Decision, as of September 30, 2013,” Statehealthfacts.org, http://kff.org/health-reform/state-indicator/state-activity-aroundexpanding-medicaid-under-the-affordable-care-act/. 4 For a good explanation of the latter changes, see Centers for Medicaid and CHIP Services (CMCS). April 19, 2012. Medicaid and CHIP in 2014: A Seamless Path to Affordable Coverage: Application, Verification and Renewals. http://www.medicaid.gov/State-Resource-Center/Downloads/4-19-12-Application-VerificationRenewals-Slides.pdf (Medicaid and CHIP in 2014). 5 CMS. “Federal Funding for Medicaid Eligibility Determination and Enrollment Activities” final rule, published in the April 19, 2011 Federal Register (76 FR 21950). See also CMS. Guidance for Exchange and Medicaid Information Technology (IT) Systems, Version 2.0, May 2011, http://www.cms.gov/Medicaid-InformationTechnology-MIT/Downloads/exchangemedicaiditguidance.pdf (accessed October 16, 2011). 6 This panel runs through December 2013, but respondents drop out over time, making the first years of data most useful. 7 Authors’ calculations, Bureau of Economic Analysis. “Personal Income and Its Disposition (Seasonally Adjusted at Annual Rates).” June 26, 2013. 8 also CMS. 2013. Verification Plan Template - Guidance and Instructions, “Phase I – MAGI-based Eligibility.” (Verification Plan Template) 9 Stan Dorn, Laura Wheaton, and Paul Johnson. “Using SNAP Receipt to Establish, Verify, and Renew Medicaid Eligibility.” prepared by the Urban Institute for the California HealthCare Foundation, April 2013, http://www.urban.org/publications/412808.html. 10 Sheila Hoag and Adam Swinburn. “Case Study of Oklahoma’s SoonerCare Online Enrollment System: Final Report.” CHIPRA Express Lane Eligibility Evaluation. Prepared by Mathematica for HHS/ASPE. May 7, 2013. http://www.okhca.org/WorkArea/linkit.aspx?LinkIdentifier=id&ItemID=14899&libID=13882. 11 42 CFR 435.952. 12 CMCS. Medicaid and CHIP in 2014. 13 42 CFR 435.952(c)(1). 14 42 CFR 435.952 (c)(2). 15 CMCS. April 19, 2012. “Webinar Series Transcript: Application, Verification, and Renewals.” http://www.medicaid.gov/State-Resource-Center/Downloads/4-19-12-Eligibility-Webinar-Transcriptpdf.pdf. 16 42 C.F.R. 435.945(j). See also CMS. 2013. Verification Plan Template - Guidance and Instructions, “Phase I – MAGI-based Eligibility.” (Verification Plan Template) 17 As CMS explained in March 2012, “States that follow procedures outlined in the regulations will not be cited for a PERM error for lack of further documentation. … PERM regulations issued in 2010 provide that PERM will 40 measure errors relative to the State’s own policies and procedures as long as those policies and procedures are consistent with Federal policy and regulations.” CMS. “Medicaid Program; Eligibility Changes Under the Affordable Care Act of 2010.” Federal Register, Vol. 77, No. 57, Friday, March 23, 2012, 17144, 17181 (Medicaid final rule). 18 CMS. “Medicaid Program; Eligibility Changes,” at 17175, 17179. In its responses to comments on proposed regulations, CMS stated the following: “We do not believe it is possible or preferable for the Secretary to prescribe all the situations in which financial data sources are useful and believe that States are in the best position to make such a determination. States currently use wage data that lags behind in making eligibility determinations and the data often is sufficient, notwithstanding the time lag, for the State to confirm the information provided by the applicant. … The time lag in the availability of quarterly wage data would not justify a State concluding that such data is not useful to verifying income eligibility and routinely relying instead on documentation provided by the individual. Conversely, a State could determine that accessing quarterly wage data is not useful if income data received from the IRS is reasonably compatible with information provided by the individual. In that situation, the agency would have obtained reliable verification of income. …. Comment: A few commenters recommended that States should not be permitted to ask individuals for additional information if the State’s data match that triggered the apparent incompatibility is more than 90 days old. Response: Data that is more than 90 days old (such as IRS data) may be relied upon to verify eligibility criteria if reasonably compatible with an individual’s attestation.” 19 CMS. 2013. Verification Plan Template - Guidance and Instructions, “Phase I – MAGI-based Eligibility.” 20 42 CFR 435.916(a)(2). 21 Final Medicaid regulations specifically require administrative renewal for Medicaid eligibility categories that use MAGI—namely, coverage for children, parents, pregnant women, and certain other non-elderly and non-disabled adults— but states may also apply administrative renewal to additional Medicaid groups. Even before the ACA, Medicaid regulations required states to make “ex parte” redeterminations of eligibility, whenever possible. In other words, renewal was required, without asking anything from the enrollee, whenever information available to the state demonstrated continuing eligibility. See, e.g., Health Care Financing Administration. April 7, 2000, and April 22, 1997, Dear State Medicaid Director letters, http://downloads.cms.gov/cmsgov/archiveddownloads/SMDL/downloads/smd040700.pdf, http://downloads.cms.gov/cmsgov/archiveddownloads/SMDL/downloads/SMD042297.pdf. These requirements were not always followed, however. 22 42 CFR §435.916(a)(3). 23 Probably the best- known example involves 401(k) retirement savings accounts. If a new employee must complete a form to open up such an account, only 33 percent do so. In that case, the default is non-participation, and most do not join. By contrast, if the default is participation, and a new account is opened unless a new employee completes a form to opt out, 90 percent are enrolled. David Laibson. “Impatience and Savings,” National Bureau of Economic Research Reporter, Fall 2005, pp. 6-8. 24 For example, one study found that, 12.6 percent of eligible Medicaid and CHIP children lose coverage each year Benjamin D. Sommers. 2005. “From Medicaid to Uninsured: Drop-Out among Children in Public Insurance Programs.” Health Services Research 40(1): 59-78. Another study found that, among children who were uninsured in 2008 but qualified for Medicaid or CHIP, 35.5 percent had received Medicaid or CHIP the previous year but were no longer enrolled. Benjamin Sommers. 2010. “Enrolling Eligible Children in Medicaid and CHIP: A Research Update.” Health Affairs 29(7):1350–1355. A study examining adult Medicaid beneficiaries concluded that, among adults who are in unusually stable eligibility categories that make them “always eligible,” 29 percent lose Medicaid each year; 43 percent of Medicaid adults lose coverage each year, with no statistically significant difference between those whose incomes grow and those whose incomes fall; and over a two-year period, 53 percent of eligible adult beneficiaries whose incomes decline nevertheless lose Medicaid. Benjamin D. Sommers. 2008. “Loss of Health Insurance Among Non-elderly Adults in Medicaid.” Journal of General Internal Medicine 24(1):1–7. 25 42 CFR 435.916(a)(2). 26 Stan Dorn and Matthew Buettgens. “Administrative Renewal, Accuracy of Redetermination Outcomes, and Administrative Costs,” prepared by the Urban Institute for the California HealthCare Foundation, October 2013. 27 CMS. “Facilitating Medicaid and CHIP Enrollment and Renewal in 2014,” SHO #13-003, ACA #26, May 17, 2013. http://www.medicaid.gov/Federal-Policy-Guidance/downloads/SHO-13-003.pdf. This State Health Official letter cited Dorn, Wheaton, et al., who made state-specific findings about eligibility levels as described in the text. 41 28 Tricia Brooks and Jessica Kendall. July 2012. “Consumer Assistance in the Digital Age: New Tools to Help People Enroll in Medicaid, CHIP and Exchanges.” Prepared by the Georgetown Center for Children and Families and Enroll America for the Robert Wood Johnson Foundation. 29 Ruth Kennedy, personal communication describing December 2009 renewals. 30 CMS. “FY 2011 Louisiana Medicaid: Payment Error Rate Measurement (PERM) Cycle 3 Summary Report.” November 31, 2012. http://ccf.georgetown.edu/wp-content/uploads/2013/01/LA.Error_.Rates_.Super_.Low_.pdf. 31 42 C.F.R. § 435.916(e). 32 42 C.F.R. § 435.916(a)(3)(ii), cross referencing 42 C.F.R. § 435.952. 33 42 C.F.R. § 435.916(a)(3)(B), cross-referencing 42 C.F.R. § 435.907(a). 34 42 C.F.R. 435.603(h). 35 CMCS Webinar Series: “MAGI Methods and Household Scenarios,” March 29, 2012. See pages 15, 20-21. http://medicaid.gov/State-Resource-Center/Downloads/3-29-12-Eligibility-Webinar-Transcript.pdf. States also have the option, via state plan amendment, to provide children with up to 12 months of continuous eligibility. Once eligibility is established for the month of application or renewal, it continues for 12 months, regardless of changes in household circumstances during that 12-month period. By contrast, continuous eligibility for adults requires special federal permission comparable to a waiver under Section 1115 of the Social Security Act. That said, basing Medicaid eligibility on projected annual income, which is allowed without a waiver, secures some of the gains of continuous eligibility by lessening the impact of income volatility on eligibility. 36 CMS apparently believes that its final ACA Medicaid regulations continue this Section’s operation. Medicaid final rule, 77 F.R. at 17171, 17174, 17176. These regulations provide that, if information needed to determine eligibility is available through the federal data hub, a state must use that hub, as a general rule. 42 C.F.R. §§ 435.948(b), 435.949(b). However, the Secretary may approve alternative verification methods that meet applicable standards. 42 C.F.R. § 435.945(k). In addition, if certain information is not available through the federal data hub— such as more detailed or recent data—states may go beyond the federal data hub to obtain that information. As CMS explained, “if verification of particular information is not available through the Federal data services hub, States may continue to utilize existing electronic interfaces.” Medicaid final rule, 77 F.R. at 17178. 37 Social Security Administration. “Programs and Data Exchange Systems.” Last updated May 15, 2012. http://www.ssa.gov/gix/programsAndDataExchanges.html. 38 Stan Dorn. Implementing national health reform: A five-part strategy for reaching the eligible uninsured, prepared by the Urban Institute for the Robert Wood Johnson Foundation, April 2011, http://www.urban.org/UploadedPDF/412335-Reaching-the-Eligible-Uninsured.pdf. 39 Catherine E. Livingston, IRS, personal communication. 40 Author’s calculations, IRS. Taxpayer Filing Attribute Report, April 2012, http://www.irs.gov/pub/irsutl/pub_4822_apr_2012.pdf. 41 U.S. Department of Education. DCL GEN-11-03, Enhancements to the FAFSA-IRS Data Retrieval Process, February 23, 2011, http://www.ifap.ed.gov/dpcletters/GEN1103.html. 42 IRS. “Topic 152 - Refund Information.” Last updated Sept. 13, 2012. http://www.irs.gov/taxtopics/tc152.html. 43 Personal communication, Len Finnochio, California Department of Health Care Services. 44 Some of these limitations would be overcome if state Medicaid programs gained access to the National Directory of New Hires (NDNH). NDNH includes quarterly wage records and new hires reports from employers in all states and the federal government. It also makes quarterly wage records available by the start of the following quarter. The Children’s Health Insurance Program Reauthorization Act of 2009 (CHIPRA) gave state Medicaid and CHIP programs access to NDNH data. See CHIPRA §203(d)(1), adding Social Security Act (SSA) §1942(a), crossreferencing SSA §453(i), the NDNH statute. The ACA took this same step, also providing exchanges with access to NDNH in determining eligibility for tax credits and other subsidies. ACA §§1413(c)(2)(A) and 1413(c)(3)(A)(ii), cross-referencing SSA §453(i). However, the HHS Office of Child Support Enforcement, which administers NDNH, maintains that these statutory provisions were not written in such a way as to provide access. It is possible that future legislation or administrative action may provide Medicaid programs with access to NDNH data, but such action has not yet occurred. 45 Many of these limitations would be overcome if state Medicaid programs gained access to the National Directory of New Hires (NDNH). NDNH includes quarterly wage records and new hires reports from employers in all states and the federal government. It also makes quarterly wage records available by the start of the following quarter. Both the Children’s Health Insurance Program Reauthorization Act of 2009 (CHIPRA) and the ACA provide that state Medicaid and CHIP agencies can obtain NDNH data to verify eligibility. See CHIPRA §203(d)(1), adding Social Security Act (SSA) §1942(a), cross-referencing SSA §453(i), the NDNH statute; ACA §§1413(c)(2)(A) and 42 1413(c)(3)(A)(ii), cross-referencing SSA §453(i). However, the HHS Office of Child Support Enforcement, which administers NDNH, maintains that these statutory provisions were not written in such a way as to provide access. It is possible that future legislation or administrative action may provide Medicaid programs with access to NDNH data, but such action has not yet occurred. 46 Personal communication, Judith Arnold. Director of the Division of Coverage and Enrollment in the New York State Department of Health; Heather Howard. Former Medicaid Director for the State of New Jersey. 47 See the discussion of “seam bias,” in the Methodological Appendix. 48 This avoids what could have been seen as a highly problematic result if a 90 percent likelihood of were required for administrative renewal with current monthly income rather than projected annual income furnishing the basis for redetermining eligibility. Such a requirement has two problems. First, whether an eligible consumer has a 55 percent chance of being administratively renewed or a roughly 10 percent change of being administratively renewed depends on a random factor—namely, whether the redetermination takes place before May 1. A consumer whose eligibility is being redetermined in January through April could be administrative renewed with prior-year income at or below 80 percent FPL, since such income would establish a 91 percent likelihood of Medicaid-level monthly income during January through April (table 5). Altogether, this could administratively renew eligibility for 55 percent of all eligible consumers (table 6). After April, however, very different results would follow. During May through August, administrative renewal would be limited to consumers who received no prior-year income and who did not begin a new job during the current year before May 1; such consumers have a 92 percent likelihood of monthly income at Medicaid levels during May through August (table 7), and they comprise only 11 percent of all eligible consumers (table 8). During September through December, the only group receiving administrative renewal would be people with no prior-year income, no current-year new job before 9/1, and current-year wages during January through April at or below 80 percent FPL. 95 percent of this group have monthly incomes at Medicaid levels during September through December (table 9), but they comprise just 10 percent of eligible consumers (table 10). Second, these results would be entirely avoided if the threshold for administrative renewal were lowered just three percentage points, from 90 percent to 87 percent. Table 4 shows the results of requiring an 87 percent or greater likelihood of eligibility. The percentage of eligible consumers who can be administratively renewed is 73 percent in January through April, 55 percent in May through August, and 45 percent in September through December. 49 Orange is available for download at http://orange.biolab.si/. 50 J.C Ham, X. Li, and L. Shore-Sheppard. 2009. “Seam Bias, Multiple-State, Multiple-Spell Duration Models and the Employment Dynamics of Disadvantaged Women.” National Bureau of Economic Research Working Paper No. 15151; Pischke, S. 1995. Individual Income, Incomplete Information, and Aggregate Consumption. Econometrica 63(4), 805-840. 43