Michael Nguyen-Mason

Dissertation Title: "Distributional Causes and Consequences of Medical Innovation"

In this dissertation I examine how the benefits and costs of medical innovation are distributed across the population, identifying the channels through which inequality in innovation operates and providing empirical frameworks for evaluating distributional equity in health markets. Across three independent papers I trace this theme from the upstream allocation of capital that finances new medical technologies, through the downstream prices patients and employers pay for healthcare, to the ultimate health gains that innovation delivers to different demographic groups. Together, the chapters characterize both the distributional causes of medical innovation—who decides which technologies get developed—and its distributional consequences—who bears the costs and captures the health benefits.

In Chapter 1 I ask whether gender bias distorts venture capital allocation in biomedical startups, testing for profit-forgoing investment decisions using comprehensive data on U.S. biomedical VC investments from 2005 to 2024. Exploiting within-city, over-time variation in the gender composition of general partners, I show that a one standard deviation increase in the share of female GPs is associated with a 22 percent increase in the share of funded founders who are women in first-round investments. However, gender-discordant investments are significantly more likely to exit via IPO, implying that both men and women GPs leave money on the table when declining to fund founders of a different gender. The homophily effect disappears in follow-on rounds when information about startup quality is more readily available, a pattern consistent with biased beliefs rather than taste-based discrimination. Finally, I document that women-founded startups are substantially more likely to develop technologies addressing women's healthcare needs, suggesting that gender bias in biomedical VC may distort not only who is funded but also which biomedical problems are prioritized.

In Chapter 2 we measure index-number bias in federal healthcare price statistics as well as the distributional incidence of healthcare inflation across the employer income distribution. To do this we construct quality-adjusted healthcare price indices from the near-universe of administrative claims in the Utah All-Payer Claims Database, linked to employer payroll tax records. Using continued procedure-by-physician pairs to control for quality, we show that traditional fixed-weight methodologies overestimate true quality-adjusted healthcare price inflation by approximately 120 basis points per year, with roughly 90 basis points attributable to substitution bias and 30 basis points to entry-exit bias. Cross-firm dispersion in inflation is far larger than aggregate indices suggest where average inflation is approximately 3.2 percent compared with a standard deviation of 10.9 percentage points. We also find that the incidence of healthcare price inflation is regressive. Regressing firm-level inflation on the rank of employer average income we find that moving from the lowest- to the highest-income employer is associated with approximately 1.6 percentage points lower annual healthcare inflation.

In Chapter 3 we develop a generalizable framework for measuring distributional health gains from pharmaceutical innovation using indices denominated in quality-adjusted life years (QALYs), separating health gains captured by the intended-to-treat population (diagnosis rate) from health gains captured by the realized utilization population (treatment rate). First, we apply this framework to the case-study of statins confirming our measure matches well documented trends. Second, we aggregate across four therapeutic areas—statins, antiretrovirals, SSRIs, and insulin/GLP-1 agents. Using MEPS data from 1996 to 2015 and QALY weights from the Tufts CEA Registry, we show that income groups with nearly identical diagnosis growth can diverge by an order of magnitude in realized health gains, a gap driven entirely by differential treatment intensity rather than differential disease burden. Finally we calculate Gini coefficients across demographic groups revealing that the level of inequality experienced by a demographic group depends on which component of health gains is examined.