Causes and Consequences of Lack of Health InsuranceGaps in Our Knowledge
Catherine G. McLaughlin, Sarah E. Crow, Mary Harrington, and Hanns Kuttner
While questions related to the causes and consequences of being without health insurance are often debated in policy and research circles, relatively little research has been undertaken to develop and test theoretical and conceptual models of why observed patterns of coverage exist. Much of the standard wisdom on health insurance comes from studies that are descriptive or that present findings from regression analyses that have not been integrated into a careful methodological framework. Too little of the work takes into account challenges such as whether a cause was actually an effect.
The six chapters that make up this book were commissioned by the Economic Research Initiative on the Uninsured (ERIU) at the University of Michigan, Ann Arbor. The initiative is part of a multifaceted effort funded by the Robert Wood Johnson Foundation to increase the nation's collective understanding of the issues surrounding the lack of health insurance coverage in the United States. The goal of ERIU is to conduct and fund research that increases our understanding of the complex interactions among labor markets, health care markets, and health insurance coverage.
In an attempt to discover which questions related to the causes and consequences of health insurance have been answered with some authority and which remain unanswered, ERIU commissioned the six critical syntheses in this book. The studies were presented in July 2001 to a conference of about 60 economists. While published research articles often begin with an abbreviated review of the economics literature, critical syntheses of a body of work are less common. Until now, no one has pulled together the breadth of the literature on the lack of coverage. Many complicated connections link health status, labor market participation, and individual decisionmaking with respect to health insurance coverage. Conducting useful research and suggesting appropriate policy changes requires a more complete understanding of those connections.
The syntheses address who does not have health insurance, why they do not have health insurance, and what difference health insurance makes. Each discusses the methodological hurdles involved in researching these questions. This comprehensive and coordinated critique of the literature serves several purposes. First, it summarizes for policymakers what we know and do not know about the uninsured. Second, it provides a framework for the health policy research needed to fill the remaining gaps in our knowledge. And third, it serves as a useful primer for economists and other policy analysts who are new to the field, making it easier for them to conduct needed research and thus contribute to the debate.
This introduction presents some of the major findings and gaps from each of the chapters. It is not an attempt to summarize the chapters fully, but rather to give the reader a sense of what they offer collectively. In short: what can we say with some confidence about the uninsured, and what areas need more research?
Counting the Uninsured
Chapter 1 addresses the question, Who are the uninsured and how many are there in the United States? Pamela Farley Short adds several new insights into how the estimates of the uninsured are calculated and interpreted. In 1998, Lewis, Ellwood, and Czajka compared estimates from the Current Population Survey (CPS) and the Survey of Income and Program Participation (SIPP) and described estimates from the National Center for Health Statistics' National Health Interview Survey, the Agency for Healthcare Research and Quality's Medical Expenditure Panel Study (MEPS), the Center for Studying Health System Change's Community Tracking Study, and the Assessing the New Federalism's National Survey of America's Families. They pointed out some of the survey design and implementation issues that should be considered when counting how many people are uninsured and who they are. Chollet (2000) looked at the same six major, nationally representative surveys above and added a discussion of the role of different methodologies in influencing estimates of how many are without health insurance.
Short takes the next step, contrasting the picture of who is uninsured that emerges from cross-sectional analyses with the image that comes from looking at who is insured and uninsured over a period of time. From a longitudinal perspective, the dualism of "insured" and "uninsured" gives way to three categories: those who are always without health insurance, those who are sometimes without health insurance, and those who always have health insurance. Drawing on this more complicated picture of the uninsured, Short analyzes the extent of continuous and sporadic coverage for various population groups. Since Swartz and McBride (1990) first analyzed spells without health insurance, it has been known that over half of the new spells without insurance end within six months. Thus the most important feature of a survey is its ability to distinguish continuous lack of coverage from sporadic lack of coverage.
Estimates from the six surveys noted above identified approximately 40 million uninsured individuals in 1999, with a difference of 7 million between the smallest and largest estimates. However, as Short points out, a number of features about these commonly used data sets merit careful consideration. For example, in many of the nationally representative surveys that collect data on health insurance, there is no direct question that identifies the uninsured. Rather, the uninsured are those who do not give positive answers to questions about different types of coverage. Because data on the uninsured are residual, forgetfulness and underreporting of health insurance are likely to inflate estimates.
In addition, cross-sectional estimates understate the number of people who experience a spell without insurance. Longitudinal estimates capture all of the uninsured spells for a sample of people over time and produce higher counts of the uninsured. Moreover, counting the uninsured over longer periods (e.g., two years rather than one) produces a higher count. Because longitudinal data allow researchers to determine the length of spells, they are also better for assessing the welfare implications of being uninsured. Short's analysis challenges much of the research that relies on cross-sectional data, including repeated cross sections used as a pseudopanel (e.g., multiple years of data from the CPS). A gross count of the uninsured obscures very different social phenomena: a person who cycles back and forth between Medicaid and uninsurance is different from a person who never has health insurance. Cross-sectional surveys have no way of differentiating between them and thus blunt our ability to identify the effects of health insurance. Chapter 1 raises questions in the mind of anyone who reads research that relies on binary categories such as insured and not insured, with and without private coverage, and so on. What difference would it make if the duration of the spell without health insurance or the propensity to remain insured was known?
The failure of the literature on health insurance to take the Short critique to heart is less a failure on the part of researchers than a limitation of the data available. The workhorse of labor economics, the CPS, produces counts of persons in various insurance status categories that Short says are "essentially uninterpretable." The SIPP follows people for at most four years. The Panel Study of Income Dynamics has begun to ask regularly about health insurance, and it may eventually become the best source of insight into the time dimension of health insurance status. MEPS asks about respondents' health insurance coverage on a monthly basis, but this rich data set is largely unused by anyone outside health economics. The Health and Retirement Survey and the National Longitudinal Study of Youth offer promise for insight into longer-term dynamics in certain age groups.
While refinements can be made on the way the uninsured are counted and described, one can say with relative confidence that there is little new ground to be broken in this area. However, researchers and policymakers alike need to be educated about the limitations of these commonly used cross-sectional estimates and about the need to look beyond them in assessing the scope of the problem and the appropriate policy response.
Reasons for Being Uninsured
In chapter 2, Linda Blumberg and Len Nichols describe what we know about why the uninsured go without health insurance. They synthesize the research on whether lack of coverage is a market failure and look at the supply and demand sides of employer-sponsored insurance.
Three main reasons have been found for lack of coverage: 1) people are healthy and choose not to have insurance because they are unwilling to pay the price for insurance; 2) people want insurance but cannot get it because of insurance underwriting practices or labor market rigidities; and 3) people want an insurance product that is available but cannot afford the coverage. Each category suggests a different set of policy prescriptions. Thus the path from research to policy requires a detailed understanding of the importance of each reason. Blumberg and Nichols separate unsubstantiated claims about causes from hypothesized causes that were based on theory and proven empirically. Beyond documenting the fact that barriers to coverage do exist, the chapter explores the reasons why some individuals have low demand for coverage and what barriers they face.
For most Americans, health insurance is a benefit provided by their employer, a practice that introduces additional players and complexity to the traditional model of supply and demand. Employers decide when, to whom, and at what out-of-pocket price to offer health insurance as a benefit. Employees, in turn, decide on job offers based not only on their skill set and wages, but also on the offer of health insurance for themselves and their families. Some of the components of these relationships are understood, and others are not. Generally, we know that employers offer health insurance to recruit and retain employees, but many of the finer details about employer offers of health insurance are not fully understood.
Blumberg and Nichols find evidence that employers' offers of insurance are influenced by worker preferences for health insurance, prevailing norms for attracting different types of workers, and the price of insurance faced by the firm. Premium price, tax incentives, and characteristics of firms and workers are significant factors in explaining the probability that an employer will offer health insurance. New, part-time, and temporary workers are less likely to be eligible for insurance, but it is not yet clear why employers make some workers rather than others eligible or what types of employers have differential eligibility policies. How do worker preferences affect employer decisions about offers? How and to what extent do preferences vary across workers? How are varied preferences taken into account in decisions about offers, eligibility rules, and the sharing of premium costs? Research has not yet satisfactorily answered these important questions.
Economic theory assumes that regardless of what percentage of the premium an employee pays out of pocket, he or she ultimately pays the remaining cost in the form of a wage-benefit trade-off. This hypothesized trade-off is not well documented empirically. Blumberg and Nichols present some evidence that employees do not pay 100 percent of insurance costs from their wages, based on how employers have responded to different policies and incentives. Some evidence indicates that the trade-off is greater when the definition of job benefits is broadened to include other things besides health insurance.
Employees do respond to the amount of the out-of-pocket premium, research shows. In fact, premium contributions have a greater effect on the rate at which employees take up the offer of insurance than total premium price does. Lower-income workers are more responsive to price than higher-income workers, and take-up rates increase with salary and earnings. Out-of-pocket premiums are even more important in determining take-up rates than the health status of the worker and family members.
Many questions remain unanswered. For example, how do workers who decline the offer of insurance differ from workers not offered insurance? What proportion of the uninsured is in transition between jobs and will soon secure coverage, as opposed to those who are continuously employed yet uninsured? What is the appropriate assumption regarding the percentage of insurance costs actually paid by workers? How does this vary by type of firm and worker, and by what mechanism are wages traded for employer premium payments? How do individuals weigh the pros and cons of private versus public coverage? Research is needed to shed light on the best way to ask and answer these questions.
Labor Supply and Job Mobility
Jonathan Gruber and Bridget Madrian focus on the interaction between health insurance and labor markets in chapter 3. Building on some of their previous work (Gruber 2000; Currie and Madrian 1999), they have developed a conceptual model outlining the hypothesized effects of worker demand for health insurance coverage on decisions regarding labor force participation (entry, exit, full-time vs. part-time work, firm), worker mobility, productivity, and wages. As they note, considerable research has looked at labor supply issues, but the majority of it merely points out the correlation between labor force participation and coverage and does not establish a causal link.
The authors' careful analysis of the literature yielded a number of findings. First, health insurance matters in people's decisions to retire or to change jobs. Second, while job lock—lack of job mobility because of fear of losing coverage—has been shown to exist, very little research has been done to determine the size of that effect or its implications for either the overall well-being of workers or the efficiency of the labor market. Third, research suggests that health insurance is not a major determinant in the decisions of low-income mothers to seek a job or leave welfare, but the authors note that more study of this topic needs to be done. There is still a lively debate about whether Medicaid expansions reduce welfare participation and increase labor supply, and there is room for significant contributions in this area.
Married couples' joint decisions with respect to health insurance coverage are also examined in this chapter. Spouses (usually identified as wives in the literature) may make their labor force decisions largely on the availability of health insurance to them and their families. Although flawed—most models are based on the assumption that the husband's health insurance coverage is exogenous, which may be increasingly unrealistic—the research indicates that health insurance is important to the labor supply decisions of "secondary" workers.
The effect of health insurance on labor markets has become an abundant field of research, and more is in the offing. New areas might take on phenomena that are puzzling to economists or not well understood. One puzzle to consider is the low rate at which part-time jobs come with health insurance offers. The standard economic account of firm behavior suggests that offers of health insurance should be independent of hours worked, and yet they are not. Another is our lack of insight into how decisionmaking among married couples affects health insurance outcomes. For example, retirement is often a joint decision. Given the higher expected health costs of the near-elderly, deciding to retire prior to becoming eligible for Medicare is likely to be influenced by health insurance options.
Impact of Health Insurance
In chapter 4, Helen Levy and David Meltzer focus on the health consequences of being uninsured. It is widely assumed that lack of coverage has deleterious effects on health status. This assumption is based on two important causative relationships: first, that being insured is critically important to receiving appropriate and timely medical care, and second, that receiving appropriate and timely medical care has a significant effect on health status.
Brown, Bindman, and Lurie (1998) reviewed much of the literature on this topic published between 1966 and 1996. They point out some of the weaknesses in the methodologies used and conclude that no causative links between lack of insurance and poor health have been supported. Levy and Meltzer build on this work. They develop a conceptual framework of the links, why one may expect different relationships among insurance coverage, health care utilization, and health status outcomes for different population groups (e.g., low income, racial and ethnic minorities) and for different diseases or treatment categories (e.g., chronic conditions, acute conditions). They also discuss what evidence would be needed to assert causation versus correlation. Very little research documents an effect of health insurance on health using the highest standard of proof, and what does exist points to small gains for people with specific ailments.
While a great many studies show a correlation between health insurance and improvements in health outcomes, Levy and Meltzer conclude that very few are controlled studies that permit a rigorous analysis of the specific contribution of health insurance on health outcomes. For a full understanding of how health insurance affects health utilization patterns and overall health, one needs an understanding of the incremental contribution of health insurance as opposed to the many other factors that affect health. Even most quasi-experimental studies have limited value because of questions regarding endogeneity. Only randomized controlled trials can provide indisputable evidence about the effect of health insurance on health status. The RAND Health Insurance Experiment was such a trial, but it was not designed to assess the impact of health insurance; rather, it was designed to test the relative impact of different types of coverage arrangements. Still, it provides some evidence that health insurance leads to limited improvements in health outcomes: namely, reduced blood pressure among low-income persons with hypertension, and improved vision.
Some readers will be frustrated by Levy and Meltzer's findings and argue that they set too high a bar for what qualifies as concrete evidence. Others will maintain that researchers and social scientists must be honest about the quality of the evidence at hand. Given problems of endogeneity and multicollinearity, as well as the lack of a controlled experiment, Levy and Meltzer conclude that we cannot definitively answer the following questions: What specific contribution does health insurance make on health status, and how does this vary across subpopulations? Does the effect differ depending on length of time with coverage? Or the frequency of spells without coverage? How do the gains and losses associated with coverage options differ across various subgroups? How cost-effective is expanded health insurance compared with other means of improving health status?
Health Insurance and Vulnerable Populations
No discussion of the dynamics of health insurance in the United States would be complete without a consideration of the special issues facing individuals who, for a variety of reasons, have a greater need for coverage and greater difficulty accessing and maintaining coverage through mainstream insurance markets. The type and degree of their vulnerability are difficult to define because they arise from varied and often interrelated factors. The way in which these factors interact and influence the availability of and demand for health insurance is also complex and difficult to specify. Some types of vulnerability, such as low socioeconomic status, are likely to influence the availability of coverage. Vulnerabilities related to language and culture, immigration status, and mental or cognitive capacity may influence the supply and demand for coverage, as well as the extent to which available coverage options are accessed and sustained.
It is important to understand whether and how traditional health insurance markets in the United States address the needs and constraints of vulnerable persons. Doing so requires knowledge about the characteristics of vulnerability and vulnerable groups, as well as an understanding of how those characteristics operate and interact within economic frameworks. For example, many economic studies include race as a dummy variable in regressions, but few go beyond this to understand the pathways by which racial minorities arrive at lower rates of insurance coverage.
In chapter 5, Harold Pollack and Karl Kronebusch provide an initial framework for thinking about vulnerability and summarize the literature on health insurance coverage of selected vulnerable populations. The authors bring together health services research and economics research on the health insurance status of vulnerable populations and provide a clear idea of how and why vulnerability affects a person's health insurance coverage. Four key components of vulnerability are set forth: barriers to accessing health insurance, poverty or economic disadvantage, discrimination, and impaired ability to make decisions. The authors then identify six populations that have one or more of these components: people with low incomes, children, racial or ethnic minorities and immigrants, people with chronic disease, the near-elderly, and individuals with psychiatric or substance abuse disorders. The authors' framework helps unravel the complexity of vulnerability and gives the reader a clearer sense of where research efforts have been targeted to date.
The authors summarize evidence from a literature review and statistical analysis of the health insurance coverage of these groups. The chapter evinces the complexity of issues affecting vulnerable populations and the need for further economic inquiry into their health insurance. While a wealth of literature discusses barriers to care for vulnerable populations, little of it focuses on the causes and consequences of lack of health insurance coverage.
Structural Models
Finally, chapter 6 synthesizes the structural models developed for and the methodological issues arising from studies of the uninsured. While the other syntheses focus on a specific aspect of the causes or consequences of the lack of insurance coverage, this synthesis focuses on how research in those areas fits together. It recognizes that almost all of the important observed phenomena discussed here are part of a grand system of interactions among employers, employees, and other consumers and decisionmakers acting to achieve their own particular objectives. How can one body of research inform the others, both substantively and methodologically? What do we learn about the relevant markets and impact of policy when we put all the pieces together?
Michael Chernew and Richard Hirth explore the connections among markets and identify aspects of the interconnections that seem to need further study. They discuss four key theoretical and methodological issues yet to be resolved that limit the ability to prescribe effective policies or predict the effect of market and policy changes: 1) the appropriate price of insurance; 2) the endogeneity of workers sorting into firms and firm decisions regarding benefit packages; 3) the frequency of transitions into and out of insurance coverage and the length of spells without coverage; and 4) the impact of labor market competition on firm behavior.
Chernew and Hirth also raise fundamental questions about the areas in which economic analysis is both most problematic and least plausible to noneconomists. For example, a fundamental proposition in economics is that consumers respond to prices. It is not clear, however, what consumers perceive to be the price of health insurance. One tradition considers the true price of insurance to be the amount above the expected cost of health care services, which is the load paid to health insurers for bearing risk, paying claims, and so on. Consumers may see the price of insurance very differently. There is substantial debate among economists about the best way to measure the price of health insurance—whether it is the load or some other measure. Without a good measure of price, economists' models cannot accurately predict behavior.
Economic reasoning about who pays for health insurance often seems implausible to noneconomists. The theory of compensating differentials suggests to economists that workers, not their employers, pay for health insurance. Even if economists generally agree with the theory, the picture of how workers pay is less clear. It surely is not the case that each individual sees his or her wage offset by his or her expected medical costs. In fact, Chernew and Hirth find evidence that employees appear not to act as if health insurance offsets wages at the individual level.
The uncertainty about how workers bear costs leads to questions about how workers sort themselves into firms that offer or do not offer health insurance. Economists make assumptions about sorting that have strong implications. If sorting is perfect, workers who place a lot of value on health insurance work for firms that offer health insurance, and those who value health insurance less work for employers that do not offer health insurance. Perfect sorting makes economic analysis easier, so it is often used in economic models; but perfect sorting suggests that we should observe things that we do not. For example, there should be some firms where workers have low demand for health insurance. These firms would rationally cater to the low demand by offering a bare-bones, high-deductible health insurance plan. Yet no such health insurance plan has been observed. Many factors make perfect sorting unlikely. To name a few, the number of firms is far from infinite. Firms are unlikely to demand homogeneous labor. Fixed costs limit the number of plans that firms offer. Efforts to discern the degree of sorting have been hampered by the possibility that sorting is based on unobservable as well as observable characteristics. Is the often-used assumption that worker characteristics are exogenous to the firm a valid one? This is one of the questions left unanswered by the structural modeling literature.
While academic economists are criticized for relying too heavily in their models on a restrictive set of possibly unrealistic assumptions, models are important because they serve as guideposts. Regression analyses developed with no theoretical underpinning can generate misleading and, in some cases, harmful conclusions about relationships. For example, specifying the endogeneity of certain worker and firm characteristics would yield important information about the degree of sorting in the market. With very strong sorting, employers are primarily a "pass through" whose decisions have little impact on the prevalence and distribution of coverage in the population. With imperfect sorting, employers become active and important players, and policies aimed at influencing employer decisions to offer health insurance are likely to be more effective.
Conclusion
Together, these chapters depict our understanding of the uninsured. They demonstrate that much of what passes for wisdom about the uninsured in policy circles is not founded on strong theoretical or empirical work, and they bring to light the tension between academics and policymakers. Social scientists will always search for better answers, whereas policymakers want the best available answers. Social scientists can wait; policymakers cannot.
This tension is not new. When Richard Doll and Bradford Hill studied the association of smoking with lung cancer and heart disease in 1950, their observational work was criticized by R.A. Fisher. Fisher argued that Doll and Hill's findings did not prove causation, and even undertook an analysis of smoking among twins to examine the possibility that a genetic factor predisposes people both to smoke and to develop lung cancer. Policymaking proceeded without waiting for evidence that would meet Fisher's standard of proof.
Since randomized controlled trials will continue to be rare in health services research, this tension between the right answer and the current answer is bound to continue. Social scientists will search for the causal linkages among health, health insurance, and the labor market, and as they do so, policymakers will draw on whatever is the best current evidence from social scientists.
The overarching message of this book is that research is badly needed to understand the links among health insurance, access to care, health outcomes, the labor market, and job choices. The uninsured are a heterogeneous group of individuals who are uninsured for differing lengths of time and for a great variety of reasons. The lack of insurance may have different effects across groups and may have different ramifications for short spells versus long spells. Understanding these differences will reveal the degree to which lack of insurance is a problem in this country and what solutions are needed.
References
Brown, Margaret E., Andrew B. Bindman, and Nicole Lurie. 1998. "Monitoring the Consequences of Uninsurance: A Review of the Methodologies." Medical Care Research and Review 55(2):177-210.
Chollet, Deborah J. 2000. "A Survey of Surveys: What Does It Take to Obtain Accurate Estimates of the Uninsured?" State Coverage Initiatives 1(5-8). Washington, DC: The Alpha Center.
Currie, Janet M., and Brigitte C. Madrian. 1999. "Health, Health Insurance, and the Labor Market." In Handbook of Labor Economics, vol. 3, edited by Orley Ashenfelter and David Card (3309-3416). Amsterdam: Elsevier Science.
Gruber, Jonathan. 2000. "Health Insurance and the Labor Market." In Handbook of Health Economics, vol. 1, edited by Anthony J. Culyer and Joseph P. Newhouse (645-706). Amsterdam: Elsevier Science.
Swartz, Katherine, and Timothy D. McBride. 1990. "Spells without Health Insurance: Distributions of Durations and Their Link to Point-in-Time Estimates of the Uninsured." Inquiry 27(Fall): 281-88.
