Economics ETDs


Justin Tevie

Publication Date



Health hazards (e.g., West Nile virus and antibiotic resistance) by their nature are detrimental to the health of mankind and are a vexing problem for society. Health authorities awareness of the rising health care costs associated with these health hazards highlights the need to undertake research in these areas. This dissertation presents a series of papers on these health hazards. Chapter 2 develops a spatial filtering panel data count model to examine the factors that contributed to the high prevalence of human West Nile virus (WNV) in California and Colorado using county-level data from 2003 to 2007. An econometric analysis was performed using a random effects negative binomial model to analyze the economic (income and home foreclosures) and biological (mosquitoes) factors associated with human WNV. Tests reveal the presence of spatial autocorrelation in the dependent variable (human WNV). The presence of this phenomenon implies that WNV in neighboring counties do impact the presence of WNV in adjacent counties. Consequently, the random effects negative binomial model is augmented with a spatially-lagged dependent variable and a spatial filtering term to correct for this problem and obtain unbiased estimates of the variance. Specification tests also show that income and home foreclosures are endogenous, i.e., home foreclosures, income and human WNV counts are determined jointly. Hence an instrumental variable (IV) technique is applied to the spatial filtering and spatial lag random effects negative binomial models to obtain consistent estimates. The former model is preferred because it is parsimonious in terms of a model selection criterion. Tests of over-identification (validity tests) reveal that the excluded instruments are indeed exogenous and for that matter valid. A number of hypotheses are tested regarding the economic and biological variables. The findings indicate that West Nile virus is higher in counties characterized by a low median income, high home foreclosures and high number of mosquito breeding sites. It is recommended that counties that exhibit these economic and biological characteristics should be allocated a higher percentage of resources for surveillance and monitoring of the disease. Chapter 3 is devoted to disease mapping and presentation of the variography of the various human WNV risk measures. It employs Geographic Information Systems (GIS) mapping tools to create thematic risk or hazard maps that visually depict the predicted probabilities of human WNV and the standardized morbidity ratios. The predicted probabilities were generated from the IV spatial filtering random effects negative binomial model. The hazard maps may ultimately assist policy makers in identifying areas of high and low West Nile virus risk, allocating scarce resources, and disease etiology. Variograms are estimated using geo-statistical methods to examine the spatial structure of the various risk measures. In this regard, both isotropic and anisotropic (directional) variograms are generated using exponential and Gaussian methods. They show the presence of strong spatial patterns in observed West Nile counts and the standardized morbidity ratios, but no spatial patterns in the model residuals. This study demonstrates how econometric methods can be used concurrently with GIS tools to inform public policy on the transmission of human West Nile virus. Chapter 4 builds a dynamic bio-economic model to study the impact of animal antibiotic use on the evolution of antibiotic resistance in humans. It reveals striking similarities between the theory of exhaustible resources in economics and antibiotic resistance. Antibiotic resistance is modeled as an exhaustible resource (common pool resource) extracted (used) over time. Each time an antibiotic is used it lowers the level of the resource (antibiotic effectiveness) by a small amount and thus raises the cost of using subsequent doses of an antibiotic. This process will continue and the next dose will lower the level of the resource even further making it more costly for future use of the drug. In other words, as more and more antibiotics are used the effectiveness of the drug dwindles over time. The planner's problem is therefore to find the optimal use of antibiotics in animals and humans over time and this necessitates the use of capital-theoretic methods. Consequently, an optimal control model is developed to examine the trade-offs between current antibiotic use in humans and animals and future antibiotic effectiveness. The results reveal that antibiotics should be used in the animal industry to the point where the immediate net marginal benefit is just counterbalanced by the long-term cost in terms of dwindling drug effectiveness. The results of the simulation exercise show that antibiotic effectiveness decreases over time because of an accumulation of resistance to the drug by bacteria. Also the shadow value of antibiotic effectiveness decreases over time because of the decreasing levels of effectiveness. Sensitivity analyses show that increased use of antibiotics in the animal industry drastically reduces the level of antibiotic effectiveness and its shadow value in a given period. The results of this dissertation could assist health policy makers in the allocation of scarce resources. The findings underscore the importance of factors such as income, home foreclosures and the number of mosquito pools in the transmission of human WNV. Thematic maps of the standardized morbidity ratios and predicted probabilities provide information on areas of high and low WNV risks. The optimal control model provides an insightful perspective on how to allocate antibiotic resources between animal use and human medicine.

Degree Name


Level of Degree


Department Name

Department of Economics

First Committee Member (Chair)

Berrens, Robert

Second Committee Member

Zhang, Guoyi

Third Committee Member

Hofkin, Bruce

Project Sponsors

Robert Wood Johnson Foundation, Center for Health Policy




Health Hazards, West Nile Virus, Antibiotic Resistance, Disease Mapping

Document Type