Harvard-MIT-BU Health Economics Seminar

Spring 2026 Schedule

SUP 951

For more information:

To be put on the Health Seminar's mailing list, please email Nicole Levitsky.
For other questions, please email Steve Walker.

April 23 Abstract:

 We provide the first causal evidence on how long-run exposure to fine particulate matter (PM2.5) at contemporary ambient levels contributes to the development of chronic disease in an initially healthy population. Our analysis links detailed personnel records for roughly 150,000 U.S. Army soldiers to high-resolution annual measures of PM2.5. Because soldiers enter service exceptionally healthy and receive uniform access to care, new diagnoses reflect pollution’s physiological impact rather than preexisting conditions or differential health care access. They are also required to relocate based on the needs of the Army which provides us with quasi-random variation in initial duty-station assignments, which we use as an instrument for cumulative pollution exposure. We find that sustained PM2.5 exposure substantially increases morbidity—particularly hypertension and related circulatory conditions—with additional effects on asthma. In contrast, short-run fluctuations in pollution have little detectable effect, underscoring the cumulative nature of pollution’s health burden even though this population is exposed to pollution levels that are considered safe by current standards. These results demonstrate that even moderate long-run exposure to air pollution can erode health capital in ways that are invisible to studies focused on short-run variations or mortality alone, and they highlight the importance of regulatory standards that account for chronic exposure risks.

April 16 Abstract:

How much additional lifetime income is equivalent to an additional year of life? This marginal rate of substitution between longevity and consumption is a fundamental parameter for economics and public policy, shaping the welfare analysis of health insurance, speed limits, environmental standards, and life-extending medical treatments. Despite its importance, the parameter is difficult to estimate credibly. Revealed-preference approaches rely on populations with non-representative preferences and require people to evaluate small changes in mortality risk, a task behavioral evidence suggests humans do poorly. We estimate the Life Year Value Elasticity (LYVE), the percentage change in lifetime income an individual considers equivalent to a 1% change in longevity, using a stated-preference survey of U.S. adults on the Prolific platform. Respondents choose between randomly varied completed-life stories that differ in lifetime income, longevity, and other life-course features. This design avoids probability judgments and elicits preferences directly over entire life courses. The average LYVE is 6.7, stable across survey waves, robust to contextual changes in the stories, and correlated with longevity-related attitudes and behaviors. We document substantial heterogeneity in LYVEs across individuals: a one standard deviation increase in a longevity-preference index predicts a 40% higher LYVE, and LYVEs decline with age beyond 60. The standard quasilinear utility model, which underlies most quality-adjusted life year analysis, assumes that the dollar value of an additional life year is independent of how long or how affluently one has already lived. Our flexible estimates reject this: the marginal value of longevity rises with income and falls with age, implying that a single LYVE cannot summarize population preferences and that policies offering meaningful individual choice could yield substantial welfare gains over one-size-fits-all approaches. Mapping our estimates to the Value of a Statistical Life under reasonable assumptions yields $13 million, near the upper end of U.S. regulatory benchmarks. We validate our estimates using a pre-registered replication experiment using GPT-5.4 and find similar income-longevity preferences, suggesting the LYVE estimates reflect a stable feature of human preferences rather than a survey artifact.