July 2006
Technology or Insurance: What Is Behind the Rapid Growth of Health Spending?
Health spending in the United States has risen dramatically over the past few decades, growing from about 5 percent of gross domestic product in 1960 to over 16 percent today. Leading experts argue that medical innovation is the principal driver of health spending growth, and studies based on the widely cited RAND Health Insurance Experiment concluded that expanded health insurance coverage through the creation of Medicare did not add substantially to the growth in health spending. At a July 17 AEI conference, Professor Amy Finkelstein presented new evidence contradicting these views. She finds that the introduction of Medicare has fundamentally altered the nature of medical practice, leading to a substantial number of new hospitals in the market and increased adoption of new medical technologies. A panel of experts debated the significance of these findings and their implication for public policy.
Amy Finkelstein
Massachusetts Institute of Technology
Four salient developments in health care motivated analysis of the role of insurance in rising costs: the spread of insurance, as evidenced by an increasing proportion of covered medical expenditures; a growth in health spending, from 5 percent of gross domestic product in 1960 to 16 percent in 2004; the development of new medical technologies; and substantial improvements in health, measured in terms of both life expectancy and morbidity.
The role of insurance in contributing to the other three developments can be analyzed by looking at the impact of the introduction of Medicare in 1965, the single largest change in health insurance coverage in U.S. history. Since different states had different levels of coverage prior to 1965, corresponding increases in coverage varied greatly by region. Comparing changes in the health-care sector by state after Medicare’s introduction allowed for analysis of the effects of insurance.
Overall, such analysis shows that the impact of Medicare on health spending is over six times larger than previously found in individual-level studies, and that about half of the growth in real per capita health-care spending from 1950-90 can be explained by the spread of insurance. Medicare may have accounted for a 37 percent increase in real hospital expenditures from 1965-70 and may have had an even larger effect had the analysis been extended beyond 1970.
Previously, technological change was believed to be the major driver of increased health spending, a belief supported by the RAND Health Insurance Experiment, which found only a modest effect of health insurance on health spending. This may be explained by the fact that the RAND experiment was performed on an individual level. Market-wide changes in insurance might have disproportionately larger effects: perhaps the introduction of Medicare encouraged hospitals to take on larger fixed costs to be shared by the new population of patients, including capacity expansions and new hospital entry. The technology-based reasoning behind increases in health-care spending is not incompatible with the findings above. In fact, the two can work together to affect costs, as increased insurance creates an incentive to develop more technology.
In addition to the costs outlined above, insurance also provides the potential for benefits, including reducing the risk of paying extremely high out-of-pocket health costs and the possibility of improved health. The analysis shows that there was no discernible impact on elderly mortality rates after the introduction of Medicare, although the analysis is limited in several ways. In contrast, the study found that Medicare reduced the financial risks facing seniors. This reduction alone may cover half to three-quarters of the cost of Medicare. Overall, insurance has been responsible for a large proportion of the increase in consumption of health care, but a substantial benefit in the form of risk reduction appears to exist.
Melinda Beeuwkes Buntin
RAND Corporation
By exploiting the differences in insurance coverage among different regions of the country, Professor Finkelstein disentangled Medicare’s effects on hospital spending from the strong upward trend on spending that already characterized the decades prior to and after the introduction of Medicare.
The results of prior studies, which examined demand and elasticity at the individual level, indicate that there is a low elasticity associated with health-related goods and services. The RAND Health Insurance Experiment is considered the gold standard of this type of study. Based on estimates from that experiment, the introduction of Medicare caused a less than 6 percent increase in total hospital spending--a much lower figure than suggested by Professor Finkelstein’s paper. However, the larger estimated impact is not inconsistent with prior research investigating the effects of technology on health spending.
An important question that the paper attempts to answer is how Medicare could have had such a large effect on health spending. The author describes two components that contribute to the rise: a fixed-cost effect and a spillover effect. The fixed-cost effect occurs when more hospitals enter the market, creating new capacity to treat Medicare patients. That effect led to an 18 percent increase in hospital spending. The spillover effect occurs when new technologies become available (because of the introduction of Medicare) and are used to treat non-Medicare patients.
The paper is carefully done and offers opportunities for further investigation. Professor Finkelstein performs numerous specification checks to rule out alternative explanations for her results--the integration of Southern hospitals being a major example. She also presents compelling graphical representations of cost trends coupled with innovative explorations of how the effects were produced. In order to provide a more comprehensive analysis, the author should consider exploring the following areas: international comparisons of health-care cost increases, the persistence of geographic variations in health spending, and effects of insurance coverage on non-hospital expenditures.
There are several major implications of this study. For example, major changes in health insurance may have larger effects than expected from microeconomic studies. There are dramatic and complex feedbacks between technology and insurance coverage that need further examination. Lastly, although the study found no discernible improvement in health outcomes (as measured by a reduction in mortality) after Medicare’s introduction, it is important to note the benefits of risk-smoothing inherent in social insurance when performing a cost-benefit analysis.
Philip Ellis
Congressional Budget Office
Previous attempts to explain the rise in health spending have attributed the cause to measurable factors, such as aging or rise in income. There remains, however, a large unexplained portion that is usually ascribed to technological improvement. The paper’s robust analysis of Medicare--a national program whose implementation’s exploitability was heretofore underestimated--is a testament to Professor Finkelstein’s ingenuity and creative techniques.
Hospital spending encompassed a large percentage of total health expenditures in the 1960s and offers a useful starting point in the author’s analysis. However, it would be useful to broaden the investigation to include variations in physician spending because Medicare did not affect only hospitals. This expansion would greatly buttress the results considering the fact that one interpretation of the RAND experiment determined cost-sharing had no direct impact on the use of hospitals. That is, cost-sharing affected whether or not one went to the doctor, who then decided if a trip to the hospital was necessary.
Although Professor Finkelstein’s paper is a solid analysis, one could draw less dramatic conclusions about the impact of the spread of insurance on increases in total health spending. For example, it appears that seniors paid 10 percent lower out-of-pocket hospital expenses in the mid 1960s as a result of Medicare. Inferring a 10 percent decrease (rather than 7 percent assumed by Finkelstein) lowers the effect of Medicare on hospital spending from 52 percent to 37 percent. That figure shrinks to 12 percent if one uses arc elasticity (which is appropriate when estimating the effect of large changes in price) rather than point elasticity. This reduction demonstrates how sensitive the results are to small changes in assumptions and methods.
A final policy consideration is that although there is an important gain from risk sharing due to insurance, we could have gotten the same benefit without spending such a large sum of money.
Mark Freeland
Centers for Medicare and Medicaid Services
There are two types of studies that one could perform to assess the causes of health-spending growth. The RAND studies are the gold standard for static analyses of micro cross-sectional data and yield demand elasticities that reflect individual responses to market changes in health insurance benefits. The majority of past studies on medical market behaviors have utilized this static approach. Few economists have undertaken a dynamic assessment of aggregate time-series data. As an exception to the norm, Amy Finkelstein has examined the collective effects of health-insurance changes--specifically, the implementation of Medicare in 1966. Her analysis captures the effects of the spread of insurance and the changing regulatory environment on both demand and supply.
The conventional wisdom that technology, rather than insurance, is the foremost cause of increasing health-care costs stems primarily from studies conducted by Joseph Newhouse and his colleagues at RAND. While the analysis revealed a strong positive correlation between technology and rising health-care expenditures, the complex interactions among other explanatory factors--including demographics, rising real income, population aging, and spread of insurance--were inadvertently attributed to the technology residual. Thus, it is impossible to identify the true contribution of technology or, for that matter, insurance expansion on mushrooming health spending.
In addition, when examining market behavior, one must distinguish between the level of insurance coverage or coinsurance and the spread of insurance in general. Reductions in cost-sharing erode the medical marketplace and distort the prices and utilization of health technology by dismantling the incentives that would exist in a functioning market. The pace of medical technological progress reflects an awareness on the part of manufacturers that third-party payers will pick up the bulk of health-care costs. Preliminary work by me and my colleagues finds that approximately half of the increase in real per-capita health spending between 1990 and 2003 was caused by insurance.
In conclusion, the principal issue is not whether technology or insurance is the major driver of health spending. Rather, the real challenge is to scientifically and objectively quantify the individual contribution of a multitude of complex and intertwined factors on health-care costs over time.
AEI research assistants Jonathan Stricks and Elizabeth DuPré and AEI intern Brian Rose prepared this summary.
