Firearm Injuries and Illicit Drug Markets: An Econometric Analysis of Costs, Violence, and Spatial Dependence

Abstract

University of WashingtonAbstract Firearm Injuries and Illicit Drug Markets:An Econometric Analysis of Costs, Violence, and Spatial Dependence Ronald Dickerson Chair of the Supervisory Committee: Professor Anirban Basu Pharmaceutical Sciences Firearm violence imposes a large and deeply unequal burden on the United States, yet the evidence base for policy remains fragmented. Two gaps are especially consequential. First, medical costs of firearm injury are typically reported at national or state levels that obscure the local variation relevant for trauma-system planning and geographic payment policy. Second, the upstream market conditions associated with lethal firearm violence remain incompletely understood. This dissertation addresses both gaps by combining insurance claims, public health surveillance, and economic data with spatial and causal inference methods to improve how firearm-related harm is measured and interpreted for policy.The first aim estimates a population-calibrated, county-level surface of mean 30-day treatment cost per treated firearm injury patient in the United States from 2016 through 2024. Using longitudinal claims from the Komodo Healthcare Map (N = 10,343 treated patients), the analysis constructs survival-adjusted episode costs, corrects for uneven claims coverage using inverse-probability weighting benchmarked to the American Community Survey, and recovers stable small-area estimates through Bayesian spatiotemporal smoothing. The case-weighted national mean predicted 30-day treatment cost was $67,275 per treated patient, and the case-weighted median was $66,985. Across counties, treatment costs were regionally organized rather than randomly distributed: raw county costs were highly dispersed, annual P90/P10 ratios ranged from 2.3 to 3.4 before smoothing, and costs were strongly spatially clustered (Moran’s I, 0.583), remaining correlated across an estimated 222 km. In a sequential decomposition, adjustment for local Medicare reimbursement geography was associated with the largest single-step reduction in geographic inequality, with the mean annual county P90/P10 ratio declining from 1.86 to 1.69. Additional adjustment for payer mix and insurance-market structure reduced the ratio to 1.59, and further adjustment for demographic composition, injury burden, deprivation, local violence burden, health care capacity, and regional utilization norms reduced the ratio to 1.51 in the fully adjusted model. These findings suggest that payment geography and local financing conditions are important, but incomplete, sources of geographic variation in paid firearm injury treatment costs. The remaining variation should not be interpreted as county-level inefficiency; it may reflect unmeasured injury severity, frailty, payer contracts, referral patterns, coding intensity, or regional delivery-system organization. The second aim examined whether exogenous changes in illicit drug prices affected firearm homicide rates in six high-exposure US cities — Dallas, Houston, Los Angeles, Phoenix, San Antonio, and San Diego — observed annually from 2016 through 2023 (N = 48 city-years). Sparse Drug Enforcement Administration transaction data were converted into complete city-year price signals through a drug-specific spatiotemporal recovery model with explicit instrument separation, then aggregated into integrated, opioid, and stimulant drug-price indices. Panel instrumental-variables models used log seizures as the excluded instrument. First-stage statistics were strong across specifications (F = 28 to 520), and placebo and falsification tests detected no evidence that instrumented drug-price variation was associated with enforcement intensity, pre-existing homicide trends, or broad non-drug-market shocks. Instrumented drug prices were negatively associated with firearm homicide across all three indices, with point estimates ranging from −0.68 to −2.85 on the log firearm homicide rate. The stimulant price specification produced the clearest evidence of a negative association, while integrated and opioid price estimates were negative but less precise. Spatial robustness checks produced broadly similar point estimates but wider uncertainty for the integrated and opioid specifications. These results do not support the hypothesis that higher illicit drug prices increased firearm homicide in this high-exposure US city sample, although the design cannot identify the specific mechanism linking price changes to violence. Together, these aims provide a new empirical foundation for local trauma-system planning and geographic payment calibration, advance spatial and causal methods for recovering interpretable structure from sparse and nonrepresentative data, and connect the downstream medical costs of firearm injury to the broader economic environments in which firearm violence occurs. The dissertation treats firearm harm simultaneously as a health care system burden — organized regionally and shaped by payment geography and delivery-system capacity — and as a geographically structured consequence of upstream economic and institutional environments.