Pamela Farley Short* and Jacob Alex Klerman**

*Pennsylvania State University
**RAND Corporation

July 1998

Since the demise of debates on how to achieve universal health insurance coverage, Congress and the states have turned to a strategy of expanding health insurance incrementally. These efforts have included market reforms to make coverage more accessible as well as expansions of federal public subsidies for children’s health insurance and, in some states, subsidies for working families.

Expanding coverage incrementally raises a host of issues and choices for public policy, with likely differential impacts on the extent to which efforts succeed in reducing either the number or proportion of uninsured. Different approaches are likely to be more or less successful depending on the extent to which they reduce financial barriers to coverage and reach out to those who are currently uninsured.

This fourth paper, Targeting Long- and Short-Term Gaps in Health Insurance, written by Pamela Farley Short, director of the Center for Health Policy Research at Pennsylvania State University, and Jacob Klerman, senior economist at the RAND Corporation, explores the types and numbers of families who would be helped by approaches to extend coverage based on different lengths of time spent without insurance. Using the federal Survey of Income and Program Participation, which tracked the insurance coverage and incomes of families over a period of nearly three years, the authors find that policies targeted toward those with the longest and shortest spells spent without insurance would reach quite different populations.

Requiring a substantial period of time spent uninsured would, in general, provide subsidies to a poorer segment of the uninsured population and cover more people than would policies targeting people with short-term insurance gaps. For example, an option that would require the applicant to have been uninsured for at least 12 months would cover an estimated 26 million people. Forty percent of these long-term uninsured individuals would be poor. An option that would require the applicant to have had insurance for at least a year and half previously, and that would offer insurance for only a six-month period, would reach 11 million people who are currently uninsured. Thirty-two percent of this short-term uninsured group would have incomes above 250 percent of the poverty level, and only 18 percent would be poor.

Twenty-six million Americans have not had health insurance for more than a year, primarily because they have low incomes and cannot afford to pay the premiums. In this report, we explore the types of families who would be helped by various approaches to extend insurance coverage based on different lengths of time spent without insurance. We also estimate the number of uninsured families who would be covered under specific options. Using the federal Survey of Income and Program Participation, which tracked the insurance coverage and incomes of families over a period of nearly three years, we find that policies targeted toward those with the longest and shortest spells spent without insurance would reach quite different populations.

Requiring a substantial period of time spent uninsured would, in general, provide subsidies to a poorer segment of the uninsured population and cover more people than would policies targeting people with short-term insurance gaps. An option that would require the applicant to have been uninsured for at least 12 months, for example, would cover an estimated 26 million people. Forty percent of these long-term uninsured individuals would be poor, and 78 percent would live in families with incomes below 200 percent of federal poverty standards (a two-person family with an income of less than $20,000). By contrast, only 24 percent of the uninsured who would have qualified for a short-term program of similar size (requiring three months of prior coverage, with a 12-month limit on eligibility) would be poor. Sixty-two percent would live in families with incomes below 200 percent of the poverty line.

Some states have implemented programs that target people who have met requirements for uninsured gaps of specific durations. Minnesota, for example, extends coverage to applicants who have not have been insured for the past four months or who have not been eligible for employer-subsidized coverage for the past 18 months. In Vermont, applicants qualify if they have not had insurance for 12 months prior to application.

Other proposed programs, which would cover unemployed individuals, typically offer only short-term coverage and require the person to have had prior coverage. A program that would require the applicant to have had insurance for at least a year and half previously, and that would offer insurance for only a six-month period, would reach 11 million people who would be currently uninsured. This short-term uninsured group has generally higher incomes than the long-term uninsured group: 32 percent have incomes above 250 percent of the poverty level, and only 18 percent are poor. In this report, we specify several stylized examples of programs that target the short- and long-term uninsured and simulate eligibility for each. For the short-term approach, we consider programs with time limits of six months and 12 months, and the examples we simulate require either three months or 18 months of prior coverage. In total, we consider four short-term programs, involving two time limits and two prior coverage requirements. For the long-term approach, we consider programs that cover the uninsured only after waiting periods of six months, 12 months, and 18 months.

Across the programs we simulate, those for the long-term uninsured cover more people and a larger share of uninsured person-years. The largest, most broadly conceived short-term program that we considered (requiring only three months of prior coverage and offering 12 months of eligibility) would have covered about 23 million people during the year and 11 million person-years. The smallest, most narrowly targeted long-term program (requiring an 18-month waiting period) would have covered about 18 million people during the year and 14 million uninsured person-years. Thus, even this smallest long-term program covers almost as many people and more person-months as the largest short-term program.

These analyses indicate that targeting the long-term uninsured is likely to both cover more of the uninsured and to cost more than targeting the short-term uninsured. Targeting the long-term uninsured has the virtue of directing public assistance to uninsured people who, given their low incomes and history of being without insurance, are unlikely to obtain insurance on their own. Targeting the long-term uninsured, however, presents other difficulties. Eligibility rules that are based on lack of insurance for a specified time are difficult to enforce. Such rules also run counter to the principles of insurance by allowing and encouraging people to remain uninsured until they have a need for medical care.

Short-term programs, on the other hand, encourage people to buy insurance and protect the insured against lapses in coverage. However, many of the uninsured who would qualify for short-term programs are not poor, and many people in these circumstances would likely find ways to continue their coverage. Accordingly, short-term programs are likely to replace—not augment—existing private insurance arrangements. Much of the money for such programs is likely to provide financial relief for the insured rather than offering new coverage for the uninsured.

Most Americans—about 85 percent—have health insurance. Their insurance is provided through a complex and interlocking set of arrangements that involves employers, government-sponsored programs, and individually purchased coverage. As demonstrated by the failed effort to enact comprehensive health reforms, voters and their representatives are reluctant to make wholesale changes in a very complicated system that serves most Americans fairly well.

The enactment of a limited set of private health insurance reforms in the Health Insurance Portability and Accountability Act of 1996 (HIPAA) and the creation of a new federal program to cover uninsured children in 1997 indicates the current political inclination to proceed a step at a time in extending coverage to the uninsured. Because of the complexity of arrangements for covering the insured, and the many reasons why some people lack health insurance, incremental reform is likely to involve a variety of approaches, targeting different subgroups of the uninsured with different types of assistance.

As we discuss in detail in the next section, pilot projects, reform proposals, and recent legislation at the state and federal levels have targeted subgroups of the uninsured according to the length of time that they have been uninsured. Some proposals target short-term gaps in health insurance, covering people who have recently lost their insurance or who would have lost their insurance in the absence of reform. Other proposals target long-term gaps in insurance, covering people who have been uninsured for some time. In this report, we contrast the two approaches and show that they reach target populations with different demographic and income profiles.

The second section of this report provides an overview of the issues, describing the uninsured according to their time without coverage and reviewing examples of programs that either explicitly or implicitly base eligibility on this criterion. The third section discusses our methodology and data source, the Survey of Income and Program Participation (SIPP). The fourth section presents our estimates of the proportion of the uninsured who would be eligible for stylized examples of programs targeting the short- and long-term uninsured, and a comparison of the target populations. The concluding section summarizes the findings and considers the implications for incremental health insurance reform.

The concepts involved in targeting the uninsured by duration of uninsured status are inherently dynamic and can be confusing, so we begin with a short review of the measurement issues. The simplest way of looking at the uninsured in a dynamic context is to focus on the stock of uninsured individuals at a given point in time.

Aggregating 12 monthly cross-sections over a year produces an estimate of the number of person-months that are uninsured or eligible for a program during a year, which is often the accounting period of interest. If no seasonal changes or other trends occur, so that the proportions of uninsureds or eligibles for the program are identical in each cross-section, then any one cross-section indicates the aggregate proportion of person-months that are uninsured or eligible over the longer accounting period. For this reason, cross-sectional estimates of eligibility and annual person-months of eligibility are closely related.

Person-months, in turn, are closely related to annual program costs. Multiplying the average cost of insuring someone for a month by the total number of insured person-months yields an estimate of total costs. For this reason, estimates of person-months of eligibility are quite useful in the evaluation of alternative policy approaches.

Programs That Target Long-Term Gaps
The long-term uninsured have also been targeted in recent legislation and proposals for reform. This approach is most prominent in states’ subsidized insurance programs. A recent survey of 16 state-subsidized insurance programs by the Alpha Center (Lipson and Schrodel, 1996) notes that nearly all state programs exclude individuals who are insured at the time of application so that people cannot drop their current coverage in order to qualify for subsidized coverage. Nearly half of the 16 state programs further restrict eligibility to those who have been uninsured for a minimum period of time. Some states even specify that the uninsured may not have access to employer-sponsored insurance for the specified waiting period. Other states have effectively imposed a waiting period by restricting eligibility to those who were uninsured on a date prior to the enactment of the legislation. Examples of these state requirements include:

• New Hampshire’s Healthy Kids program, which excludes children who have been enrolled in an employer plan within the past three months.

• Minnesota’s MinnesotaCare, which requires individuals to have been uninsured for four months and to have had no access to coverage for which an employer would pay at least half.

• Vermont’s Health Access Plan program. which excludes people with insurance in the preceding 12 months (although this requirement may be waived under special circumstances).

• New Jersey’s Health Access program, which excludes people who had employer-based insurance in the last 12 months.

• Oregon’s Family Health Insurance Assistance Program, which requires individuals to have been uninsured for 12 months prior to application.

• New York’s Regional Pilot Projects, which require individuals to have been uninsured on some earlier date.

• Tennessee’s TennCare, which requires individuals to have been uninsured on some earlier date.

Similar restrictions were included in 11 state demonstration programs involving employers who did not offer insurance that were sponsored by the Robert Wood Johnson Foundation. These demonstrations were targeted toward the "chronically uninsured" according to Helms, Gauthier, and Champion (1992). All but one were restricted to employers who had not offered health insurance for a minimum of six months to a year.

III. METHODS AND DATA

In each instance, the simulation proceeds by looking back in time from each month in which a person is uninsured to determine whether the person meets the eligibility criteria for each of the short- and long-term programs. Having made this calculation for each month in the one-year period, we estimate how many uninsured people would be covered by the program at some point in the year and how many uninsured person-months would be covered.

We also characterize the covered people and person-months in terms of their percent distribution by family income, age (adult or child under age 19), family type, and family employment. For these purposes, "family" refers to a "health insurance unit" (a married couple or single adult and dependent children). This nuclear family concept corresponds to the group of people who would typically qualify for family coverage under a single private insurance policy.

We characterize people according to their status at the start of the year for our estimates of those who were ever uninsured or eligible for one of the model programs over a year. We use the monthly data for estimates involving person-months for these characteristics.

The data source for these simulations and tabulations is the 1991 panel of the Survey of Income and Program Participation (SIPP). SIPP is a panel survey, conducted by the U.S. Bureau of the Census, that follows a nationally representative sample of the U.S. population over a period of nearly three years. At every interview, the survey collects information on monthly health insurance status (including source of coverage) and employment status during the preceding four months. In addition, a supplement to the second interview collects retrospective information on how long each insured person had been covered at the start of the panel and the last time that each uninsured person was insured.

For our analysis, the crucial variable is the length of time that a person has been uninsured. In mimicking some proposals, we also consider length of enrollment in private insurance at the time someone became uninsured. Because retrospective recall of such durations is known to be less than perfect, we base our estimates as much as possible on durations that are captured within SIPP by tracking respondents over time. More specifically, we make estimates for people who were uninsured in each of the last 12 months of the panel so that we can use the data collected at each of the earlier interviews to calculate the prior duration of uninsured or privately insured spells.

IV. FINDINGS

Within the range of program designs for which there is precedent, the programs that target long-term gaps in health insurance provide more person-months of coverage over a fixed accounting period (such as a year) than programs that target short-term gaps in coverage. Given the link between subsidy costs and person-months of coverage, subsidies for the long-term programs also cost more. The differences between the short- and long-term approaches are somewhat smaller when measured in terms of the number of people who would benefit over time. However, as a general rule, the long-term programs will cover more people as well.

As measures of the uninsured "gap" and program eligibility, both people and person-years have policy relevance. Counts of people show how many lives are affected by lack of insurance and could benefit from an incremental expansion of coverage. Because of the link between costs and person-years, counts of person-years come close to measuring costs and benefits in accounting terms.

Our results also show that programs targeting short- and long-term gaps in health insurance would cover different segments of the uninsured population. In particular, the long-term uninsured tend to be from low income families. The short-term uninsured (who recently had private insurance to lose) tend to be from higher income families. The composition of these groups differs as well.

Our findings indicate that many of the uninsured who would benefit from short-term programs are people who are not poor and, in some cases, are well above the poverty line. Consequently, it seems reasonable to expect many of the short-term uninsured to pay for their own coverage. That was apparently the logic behind both COBRA and HIPAA, programs that targeted short-term gaps but did not offer subsidies.

Many people in the same circumstances—for example, people who have lost a job that offered employer-sponsored insurance—do indeed pay for continued coverage (Klerman 1997). That observation not only raises the question of whether subsidies are really necessary to close the remaining short-term gaps in coverage, it also means that a broad subsidy program to cover the short-term uninsured would assist many people who would have obtained and paid for health insurance anyway. This is a significant reason to steer away from subsidies for the short-term uninsured.

Implicit in the current emphasis on targeted health care reforms is the policy goal of reducing the number of uninsured with a modest investment of public money and political capital. Offering subsidies to the short-term uninsured is unlikely to meet this goal because of the likelihood that many of the subsidies would go to people who would have obtained insurance through private means. Subsidies to such people give them financial relief but do not reduce the number of uninsured.

While subsidies may not be the best way to close the gap for many of the short-term uninsured, directing some time-limited subsidies to low income individuals and families who lose their insurance might be necessary to encourage them to obtain insurance. For the rest of the short-term uninsured, other ways could help make insurance more affordable. For example, unemployment insurance might be redesigned to allow workers to insure against the cost of health insurance during periods of unemployment. And the development of less expensive alternatives to employer-sponsored insurance that individuals could buy on their own would also help to fill and avoid short-term lapses in coverage.

Nearly by definition, the long-term uninsured are unlikely to obtain private insurance. By satisfying a long waiting period requirement, people reveal that they have little ability or desire to obtain insurance—a reflection of some combination of lack of access, high premiums, low incomes, or little perceived need for insurance. Empirically as well, hazard modeling of the probability of ending an uninsured spell suggests that people who are uninsured for more than a year are quite unlikely to acquire insurance in the near future (Swartz and McBride, 1990). Consequently, targeting subsidies at the long-term uninsured has the virtue of avoiding a significant displacement of private insurance. Given the income distribution of the long-term uninsured, offering significant subsidies to encourage a large proportion of these people to obtain insurance would likely be necessary.

Targeting the long-term uninsured, however, presents its own difficulties. For one, rules that base eligibility for subsidies on lack of insurance for a specified time are difficult to enforce. Lipson and Schrodel (1996) report that Florida’s Healthy Kids program had a waiting period (six months without health insurance), but dropped it because of enforcement difficulties.

The reverse case, in which eligibility depends on establishing a history of insurance, offers an instructive contrast. HIPAA grants a variety of continuation and portability rights to former employees with a minimum history of continuous coverage. To identify employees who qualify for these rights, the legislation requires employers to provide written certificates of prior coverage.

How would one develop a similar certification program to establish how long a person has been uninsured? No single institution has the information to certify this information. A national registration system for health insurance coverage will not likely be established merely to enforce the waiting period for a subsidy program. Thus, long-term programs are left to fall back on requiring people to swear or sign that they are eligible. That enforcement mechanism is obviously weak and prone to fraud.

Requiring long periods without insurance also raises a moral issue. Program administrators in Washington and Florida noted that the "right thing" for the uninsured to do is to buy insurance for themselves in the private market (Gauthier and Schrodel, 1997). Requiring participants to be uninsured for a minimum length of time penalizes people who did the right thing by being prudent and buying insurance out of their own pockets. This rule also encourages people to do the wrong thing: remain uninsured in order to satisfy the waiting period requirement.

Put in another way, in an ideal world consumers would buy health insurance policies that would cover them for a lifetime (Cochrane 1995). Lifetime insurance contracts, which obligated purchasers to pay premiums and insurers to pay claims over a lifetime, would protect consumers against the risk of changes in health status that make them "uninsurable." Lifetime insurance contracts would also hold down the cost of health insurance by preventing people from waiting to buy insurance until they were ready to file a significant claim. Coverage programs with long waiting periods encourage the latter sort of behavior and move away from the ideal. Short-term programs, especially those that target people who have already paid for insurance for a significant time, move closer to the ideal. But short-term programs are also the most likely to replace, and not augment, existing insurance arrangements.

As this discussion demonstrates, targeting either the long-term uninsured or the short-term uninsured has advantages and disadvantages, as does any program to extend coverage to the uninsured. Our principal intention here is to call attention to the difference in the size and composition of the groups that are targeted when incremental programs focus on short- and long-term gaps in health insurance. Programs focused on short-term gaps will usually reach fewer of the uninsured, will particularly benefit people with middle and high incomes, and should probably emphasize strategies other than subsidies to encourage more health insurance purchases. Programs focused on long-term gaps will usually cover more of the uninsured, will particularly benefit people at the bottom of the income distribution range, and for that reason will almost inevitably require public financing for many of the newly insured.