Date of Award

Spring 1-1-2018

Document Type


Degree Name

Doctor of Philosophy (PhD)

First Advisor

Daniel Kaffine

Second Advisor

Scott Savage

Third Advisor

Chrystie Burr

Fourth Advisor

Oleg Baranov

Fifth Advisor

Ryan Lewis


My dissertation examines several theoretical and empirical issues under the context of non-private goods that experience imperfect market mechanisms. Each successive chapter covers the respective topics of the provision of public goods through procurement auctions, the monopolistic competition of digital goods exhibiting network externality, and the depletion of common resources with poorly defined and seldom enforced property rights.

In the first chapter, I estimate the structural parameters of both uncertainty and its heterogeneity with respect to firm size in highway procurement auctions within a semiparametric generalized method of moments framework. The estimation results allow further analyses of firm behavior and auction design through calibration and counterfactuals. In addition, the paper shows that structural parameters can be extracted from heteroskedasticity under fairly simple assumptions, and the method may be extended to the study of other market settings with heteroskedastic outcomes.

In the second chapter, I develop a symmetric, static circular city model of monopolistic competition where firms supply differentiated network goods with a freemium strategy and heterogeneous consumers who can adopt multiple products at an increasing marginal cost. The model shows that freemium competition in the static, symmetric setting may not be supported if marginal cost to users or firms is too high, and free-entry may have a deleterious effect on social welfare. However, with sufficiently high fixed cost, a competitive social optimum can be achieved. The model establishes a flexible framework for analyzing network competition with freemium strategy and multiple adoption and is shown to be easily adaptable to various extensions.

In the third chapter, I develop a class of dynamic land-use models with continuous outcome accounting for both intensive and extensive deforestation and, as a proof of concept, incorporate the Kalman filter to improve regression-based spatiotemporal land use forecast under simplifying assumptions. Using remote sensing data of forest cover in the Amazon Rainforest, I show that even with rudimentary specifications and minimal explanatory variables, Kalman filter combined with an economic structure yields superior predictive power than regression alone, and I further propose a more robust framework for future implementation.