Loading field map
Why do people spend money on doctor visits, gym memberships, and sleep instead of spending all their income on consumption goods? The answer seems obvious—health matters—but for economists trained to model choices between competing wants, health posed a puzzle. Health is not a typical good that you buy and consume immediately. It is a stock that depreciates, can be built up through investment, and affects how much time you have to earn income and enjoy life. The subfield of health capital and demand for health grew out of the effort to turn that intuition into rigorous economic models, and the frameworks that emerged have shaped how economists think about medical care, insurance, and health behavior ever since.
The foundational framework for this subfield is the Grossman Health Capital Model, introduced by Michael Grossman in 1972. Grossman proposed that health should be understood as a durable capital stock. Individuals are born with an initial endowment of health capital, which depreciates with age and can be augmented through investments such as medical care, diet, and exercise. The model gives health a dual role: it is both a consumption good—people value feeling well—and an investment good—better health means more healthy days available for work and leisure. This investment motive was the radical move. By treating health as a form of human capital, Grossman reframed medical care as a derived demand: people do not want doctor visits for their own sake; they want the health that doctor visits produce.
The Grossman model is built around an intertemporal optimization problem. Individuals maximize lifetime utility subject to a budget constraint and a health production function. The key result is that the optimal stock of health is determined by equating the marginal cost of health investment with the marginal benefit of an additional healthy day, discounted over the remaining lifetime. The model generates predictions about how health investment responds to changes in wages, education, age, and the price of medical care. For example, higher wages increase the opportunity cost of sick time, so the model predicts greater health investment among higher earners. Education is predicted to improve the efficiency of health production, so more educated people get more health from the same medical inputs.
The Grossman model remains the theoretical backbone of the subfield. It gave economists a language for talking about health as something people actively produce rather than passively receive. But its very ambition—a full intertemporal optimization framework—also created tensions that later frameworks would address.
At almost the same moment that Grossman published his theoretical model, a separate line of work emerged that focused more narrowly on the demand for medical care itself. The Demand for Medical Care Framework, also dating to the early 1970s, was less concerned with the deep structure of health capital and more concerned with estimating how people respond to prices, insurance coverage, and access. The landmark RAND Health Insurance Experiment (1974–1982) epitomized this approach: it randomly assigned families to insurance plans with different cost-sharing levels and measured how their medical spending changed. The central finding—that higher cost-sharing reduces demand, but with little effect on health outcomes for the average person—became a cornerstone of health policy.
Where the Grossman model treated medical care as one input among many into a health production function, the Demand for Medical Care Framework treated medical care as the outcome of interest. This narrowing was deliberate and practical. Policymakers needed to know the price elasticity of demand for doctor visits and hospital stays to design insurance contracts, set copayments, and predict the budgetary impact of coverage expansions. The framework produced a rich empirical literature on how demand varies by type of service, by patient characteristics, and by insurance design.
Yet this practical focus came with limitations. The Demand for Medical Care Framework was largely static: it estimated responses to current prices without modeling how those responses fed back into future health. It also treated preferences as given, without exploring why some people invest more in health than others. The Grossman model offered a deeper explanation, but the demand framework offered testable predictions that could be directly applied to policy. The two frameworks coexisted uneasily—theoretical depth versus empirical tractability—and that tension would shape the next wave of methodological development.
By the 1980s, health economists had accumulated a large body of demand estimates, but they faced a credibility problem. People who use more medical care are different from those who use less—they are sicker, richer, more anxious, or better insured. Simple correlations between price and demand could be driven by these unobserved differences rather than by true price responsiveness. The Microeconometrics in Health Economics school emerged to address this endogeneity problem. Drawing on the broader credibility revolution in empirical economics, microeconometricians brought quasi-experimental methods—instrumental variables, difference-in-differences, regression discontinuity, and natural experiments—to health demand estimation.
The microeconometric approach narrowed the focus even further than the Demand for Medical Care Framework had. Instead of estimating structural parameters of a health production function, microeconometricians aimed to identify causal effects of specific policies or treatments. The RAND experiment had been a randomized trial, but most policy variation is not randomized. Microeconometrics provided tools to exploit natural experiments: a change in Medicaid eligibility that affects some states but not others, a drug price shock caused by patent expiration, or a discontinuity in insurance coverage at age 65. This school transformed health economics by making causal identification the gold standard.
What microeconometrics gained in credibility, it arguably lost in theoretical structure. The Grossman model's intertemporal optimization was largely set aside in reduced-form work. Researchers estimated the effect of a copayment change on doctor visits without modeling the underlying health capital dynamics. The microeconometric school did not reject the Grossman model—many practitioners accepted its broad logic—but it treated health demand as a series of separable empirical questions rather than as a unified optimization problem. This created a gap between theory and empirical practice that the next methodological school would try to close.
The Structural Econometric Modeling in Health school, which gained momentum in the 1990s, represented a deliberate return to the Grossman model's theoretical ambitions. Structural modelers argued that reduced-form estimates, while internally valid for a specific policy change, could not predict the effects of policies that had never been tried. To simulate counterfactual policies—a new insurance design, a different subsidy structure, a change in the retirement age—you needed a model of the decision process itself. Structural estimation involves writing down an explicit dynamic optimization problem, typically building on the Grossman framework, and using data to estimate the deep parameters of preferences, health production, and depreciation.
The disagreement between microeconometrics and structural modeling is not merely technical; it reflects different views about what economic knowledge is. Microeconometricians tend to see structural models as fragile—dependent on functional form assumptions that are hard to test. Structural modelers see reduced-form estimates as atheoretical—useful for answering "what happened" but not "what would happen if we changed the rules." This living disagreement has shaped the subfield for three decades. Some researchers have tried to bridge the gap by using structural models to generate predictions that can be tested with reduced-form methods, or by embedding quasi-experimental variation within a structural framework.
Structural modeling revived the Grossman model's original program of understanding health investment over the life cycle. It allowed researchers to study how health and wealth co-evolve, how early-life health shocks propagate into later outcomes, and how Medicare or Social Security policies affect health investment incentives. But structural models are computationally demanding and require strong assumptions about how individuals form expectations. They have not displaced microeconometrics; rather, the two schools coexist, each with its own domain of application.
While the methodological debate between reduced-form and structural approaches was unfolding, a different challenge to the Grossman model emerged from behavioral economics. The Behavioral Health Economics framework, which took shape in the 1990s and expanded rapidly after 2000, questioned the rational-actor assumptions at the core of the Grossman model. Grossman's optimizing agent has perfect foresight, exponential discounting, and stable preferences. Behavioral health economists pointed to evidence that people smoke despite knowing the risks, fail to take medications as prescribed, procrastinate on preventive care, and choose health insurance plans that are clearly dominated by cheaper alternatives.
Behavioral models introduced present bias (hyperbolic discounting), loss aversion, limited attention, and social norms into the analysis of health demand. A present-biased person might intend to exercise tomorrow but always choose leisure today, creating a gap between long-run preferences and short-run choices. This insight directly challenged the Grossman model's prediction that people invest optimally in health given their discount rates. Behavioral health economists showed that small changes in the choice environment—default enrollment in health plans, automatic refills for prescriptions, commitment contracts for gym attendance—could have large effects on health behavior, effects that a rational model would not predict.
The behavioral framework also transformed the Demand for Medical Care Framework's policy implications. If people are present-biased, then cost-sharing might reduce not only wasteful care but also valuable preventive care that patients undervalue in the moment. Behavioral economists argued for "nudges" that steer people toward better choices without restricting freedom: simplifying enrollment, providing social comparisons, or changing the order of options in a menu. These tools have been adopted by health systems and insurers, making behavioral health economics one of the most policy-relevant frameworks in the subfield.
Behavioral health economics did not replace the Grossman model; it modified it. Many current models incorporate behavioral features—present bias, reference-dependent preferences, or limited cognition—into an otherwise Grossman-style optimization framework. The behavioral critique also enriched the Demand for Medical Care Framework by explaining why demand elasticities might differ across populations and contexts in ways that standard price theory could not capture.
Today, all five frameworks remain active, but they have settled into a division of labor. The Grossman Health Capital Model continues to provide the theoretical foundation for studies of health investment over the life cycle, especially in structural work. The Demand for Medical Care Framework survives in applied policy analysis, particularly in insurance design and cost-sharing research. Microeconometrics in Health Economics is the dominant empirical toolkit for causal inference, used in thousands of studies on the effects of health policies, insurance expansions, and medical interventions. Structural Econometric Modeling in Health is a smaller but influential school, essential for policy simulation and for understanding dynamic health behaviors. Behavioral Health Economics has become a mainstream perspective, integrated into both theoretical and empirical work.
What the leading frameworks agree on is that health is not a standard consumption good and that understanding health behavior requires attention to incentives, information, and constraints. They disagree on how much of health behavior can be explained by rational optimization versus psychological biases, and on whether empirical work should prioritize internal validity (microeconometrics) or predictive power for counterfactual policies (structural modeling). These disagreements are productive: they drive methodological innovation and prevent any single framework from becoming dogmatic.
The most exciting current research combines frameworks. Structural models now incorporate behavioral features like present bias. Microeconometric studies test predictions derived from behavioral models. The Demand for Medical Care Framework has absorbed behavioral insights about how framing and defaults affect choices. The Grossman model has been extended to include social determinants of health, peer effects, and the role of early-life conditions. The subfield's history is not a story of one framework triumphing over others but of a set of intellectual tools that continue to evolve in response to each other and to the pressing policy questions of the day.