In this dissertation, I explore and empirically evaluate the management of an endemic disease in Chile's salmon aquaculture industry. Chile is the world's second-largest producer of farmed salmon and the largest supplier of salmon to the United States. However, the industry has struggled to manage endemic, transmissible diseases that threaten industry productivity. This has driven the industry to utilize antibiotics at a rate that has raised concern among public health and environmental communities.
Given the ongoing challenges from endemic pathogens, what avenues are available for balancing disease management with antibiotic stewardship? One way to address the spread of pathogens is through coordinated action between farms. In the first chapter, I evaluate a spatially explicit disease management program that forces all farm sites within the same neighborhood to coordinate their production activities. Theory suggests that this system will be most effective when transmission between neighborhoods is low because it reduces the likelihood of immediate re-infection at the start of the coordinated production cycle. I evaluate the extent to which pathogens spread between neighborhoods, exploiting exogenous variation in the timing of production cycles induced by the disease management policy. I do not find any evidence of spillover between neighborhoods, suggesting that the spatial scale of the policy is appropriately matched to the spatial scale of transmission. To account for the possibility of a behavioral response that could mitigate observed disease prevalence, I also test for changes in the propensity and intensity of antibiotic use that might be caused by the variation in pathogen pressure. However, I find no evidence of such effects.
In the second chapter, I turn toward the private incentives for disease control, specifically antibiotic use. The incentive to apply antibiotics is determined by its effectiveness at reducing losses from disease. To simulate counterfactual disease control policies, it is therefore necessary to accurately estimate the effectiveness of antibiotic treatments in a farm setting. However, several econometric challenges are associated with measuring the effectiveness of treatment using observational data. Using an epidemiological model with endogenous antibiotic applications, I illustrate how self-selection into treatment adds bias to an empirical estimate of a structural parameter critical for counterfactual simulation of disease dynamics. This bias is driven by the behavioral feedback between treatment and stochastic features in the epidemiological model that are unobserved by the econometrician. When producers self-select into treatment based on these unobserved features, conventional models tend to underestimate treatment effectiveness. I illustrate an alternative estimator based on the control function approach and conduct a Monte Carlo experiment to illustrate its efficacy. I then apply the estimator to data from the Chilean salmon farming industry to illustrate how the estimator can be used in practice.
In the final chapter, I simulate the impacts of targeted reductions to antibiotic use to explore the tradeoffs between disease management and antibiotic stewardship. To do this, I solve for the dynamically optimal antibiotic treatment schedule in the presence of disease transmission between individuals on a farm and between farms. To capture the effect of the neighborhood coordination policy, I parameterize the disease pressure from other farms in a time-varying manner consistent with coordinated production cycles. I then simulate the impacts of various antibiotic reduction policies that feature varying levels of effectiveness and expected implementation costs. Each of the restrictions causes a non-uniform shift in the timing of treatment in the production cycle, an effect that is ultimately consequential for the disease-related externality. I find that the most cost-effective instruments for private individuals also tend to generate the largest externalities.
Each chapter is supported by data obtained from the Servicio Nacional de Pesca y Acuicultura (SERNAPESCA), the national regulatory authority for Chile's aquaculture industry. These data are collected as part of a mandatory reporting requirement and provide georeferenced, weekly observations on the production status and veterinary health of all of the industry's fish farms. Critically, these data also include prescriptions for antibiotics applied for a wide range of purposes, including the treatment of Salmonid Rickettsial Syndrome (SRS), which is caused by the endemic pathogen \textit{Piscirickettsia salmonis}. Although these data have been used in past epidemiological studies, this dissertation is the first to utilize these data for economic analysis. In some cases, I pair this database with supplementary data, drawing upon publicly available information on policy borders and coastlines that are central to the empirical investigation in the first chapter.
