I study the effect of specialty credit cards and independent stations on competition in the Korean gasoline market. Chapter 1 shows how discounts from specialty credit cards work as artificial switching costs and help firms to increase price through splitting customers according to each brand in the industry. The empirical results from the Korean gasoline market in chapter 2 show that gasoline price increased after the introduction of brand-specific gasoline-discount credit cards and decreased after a partial replacement of the cards by all-around discount cards. This would harm customers without the credit cards and offset the discounts from specialty credit cards. Chapter 3 shows that independent stations set lower prices than nearby stations which in turn leads the nearby stations to lower their prices. The "rockets and feathers" effect is reversed among independent stations but is stronger among refiner operated stations. This supports the tacit collusion theory as a source of the phenomenon.

Plant-based alternatives have been gaining attention due to their health and environmental benefits, as well as their technological improvements that make the taste, texture, and appearance more similar to animal-based meats and milk. Demand estimates from this study’s nested logit model show that people are much more flexible in their milk choice than their meat choice, which is likely due to the fact that milk has a much wider array of applications than meat. The results show that consumers care about the health effects of their choice, but still have a higher willingness to pay for products that best satisfy their taste preferences. With an understanding of which characteristics consumers value in animal-based meats, research and development efforts in the plant-based protein industry can be more focused in trying to develop those attributes. Decreasing the environmental footprint associated with agriculture requires not only adopting plant-based products but also analyzing the marginal costs of the beef industry with more scrutiny. Through a random-coefficient logit model, including supply-side estimation, federal meat crop subsidies incentivize beef production because they are shown to have a small, but negative impact on marginal cost.

This dissertation approaches network effects from different perspectives and methodologies. Social network effects of immigrants have an impact on the decision of their destination, the presence of this characteristic is exploited in the first chapter to statistically isolate the effect they have on the voting behavior of the district’s representative. Generally, more immigration is correlated with a more liberal voting behavior. The second chapter takes a theoretical approach, a network adoption model is constructed that deviates from the usual linear assumption and incorporates time of adoption and a type-dependent utility function that allows a different utility derivation dependent on the type of members in the networks and not only in size as the linear model assumes. Using simulations, it is found that more heterogeneous populations and the presence of early adopters allow the emergence of clusters and the survival of small platforms even in the presence of strong network effects.

The smart lighting control system requires data of the indoor daylight to optimize the daylight utilization to reduce electrical energy consumption of lighting and simultaneously increase occupants’ comfort level and healthfulness. However, existing daylight sensing systems are often inaccurate or easily interfere with occupants’ activities. We put forward an indoor daylight sensing system based on the “one-phase” Hardware-in-the Loop (HiL) Simulation. The core of this new HiL system is a new luminance distribution sensor (referred to as Parallel Luminance Sensor, PLS) which can measure the emitted luminance distribution of the building fenestration in multiple directions simultaneously (i.e., in parallel). The PLS is designed to measure the luminance distribution in 145 discrete patches based on the Klems basis. The PLS has a dynamic range of 23 bit and a sampling rate of 10 Hz. Several laboratory and field deployments demonstrated the efficiency of the proposed and designed PLS-enabled HiL system for daylight harvesting and/or artificial light control to reduce the energy consumption. The test results showed that the system is able to estimate the daylight illuminance of a large area with only one single sensor with good accuracy. Besides, the system can generate luminance maps to be used for evaluating the visual comfort and/orglare. The aspects for improvement are also identified from the test results. The cost-benefit analysis shows that the new system can reduce the number of sensors installed and achieve savings in commissioning cost while reducing the interference between the occupants and the sensor system. The future plan towards the improvement of the new system and its real-world applications is discussed.

The first chapter of this dissertation replicates, re-specifies, and re-estimates Cardon and Hendel's 2001 RAND Journal of Economics article "Asymmetric Information in Health Insurance: Evidence from the National Medical Expenditure Survey." This article presented the first structural model to study adverse selection in the health insurance market—finding no evidence of informational asymmetries. I demonstrate, however, that after normalizing residual income, correcting an inconsistency in the construction of a health good, and accounting for corner solutions of the maximization problem, asymmetric information in fact exists in the health insurance market studied.

The second chapter examines whether hospitals strategically choose to vertically integrate with physicians in order to capture facility fees. To address this question, I match data on hospitals' ownership of clinical oncologist practices with Medicare payment data disaggregated to the physician and specific service level. I leverage a 2014 policy change that drastically altered the payment structure of Medicare's facility fees paid to hospitals for evaluation and management services—and yet, it did not alter the direct payments made to physicians. Contrary to popular belief, I find no evidence that the financial incentives of facility fees have an effect on the probability that a hospital and clinical oncologist vertically integrate.

The third chapter explores how hospital-physician integration in the United States has the potential to restrict patient referral pattern flows within specific referral networks. I first empirically illustrate these referral network restrictions that result from hospital-physician integration and then consider their implications for newly integrating oncologists as well as for independent oncologists. Utilizing detailed longitudinal data that cover the complete U.S. shared patient patterns for oncologists' Medicare beneficiaries along with oncologists' integration status with hospitals in geographically-defined market boundaries, I find that the average integrating oncologist increases his or her share of referrals made to health system partners by 36 percentage points following integration; these effects are most pronounced in markets with highly concentrated levels of integrated oncologists employed by a single health system. In addition, hospital-physician integration increases the probability of referral foreclosure—with respect to the referrals made to oncologists outside of a health system—by 14 percentage points in unconcentrated markets and by 23 percentage points in highly concentrated markets. As a single health system increases its market share of integrated oncologists, independent oncologists in the market shift their referrals to oncologists of lower quality. These findings suggest broader access issues for patients of independent oncologists in markets that experience high levels of consolidation that result from hospital-physician integration.

This dissertation consists of three essays in empirical industrial organization. It uses a combination of structural and reduced-form econometric methods utilizing unique data sets.

Auctions for sale of agricultural produce have been gaining popularity in recent years. The first essay uses a game-theoretic model of behaviour within the independent private-values paradigm (IPVP) to put structure upon a unique dataset collected from one of the seven brokerage firms that conducts auctions for sale of tea in Bangladesh. Incorporating insights from the theory of auctions, the first essay estimates the distribution of valuation for the bidders in the Chattogram tea auction, both parametrically and non-parametrically. This flexible, realistic empirical framework enables the study of the equilibrium relationship between bidding behaviour and sale prices in the tea auctions. The results from counterfactual exercises suggest that the existing mechanism of oral, ascending price auction, actually would garner higher revenue for the auctioneer compared to a Dutch auction. On the other hand, implementation of a Generalized Vickrey Auction (GVA) in case of multi-unit sale would generate substantially higher revenue than single-unit English auctions. This paper also estimates the optimal reserve price for the auctioneer as well as the average surplus for the bidders. In recent years, voters and city councils in cities throughout the U.S. have implemented taxes on sugar-sweetened beverages (SSBs). Some economists have argued that taxes on these drinks could lower consumption, particularly among high-risk populations and that lower consumption could reduce the prevalence of obesity. On the other hand, many retailers are also opposing the soda tax as they claim that such tax reduces their total sales. Since beverages are often “loss leaders”, sold cheaply to attract shoppers, by raising the price of the product the tax reduces the value of beverages as an attractive force. In contrast, stores, which are not taxed, could undercut prices and attract customers from the stores with soda tax. That would reduce demand for all products sold in the stores with soda tax, not just beverages. In the second essay, I tried to see how the soda tax has affected sales of the retail stores in Berkeley, which is the first city in the USA to implement such a tax. Using high-resolution scanner data and data-driven approaches to select comparison units for counterfactual analysis, I estimate that the tax has no significant effect on sales. The third essay develops an empirical model of consumer usage and price uncertainty under a three-part tariff plan to study the effects of an enacted ''bill shock'' agreement in mobile telecommunication markets, which requires mobile network operators to inform consumers when they use up the monthly allowance of their mobile phone plan. Using a rich billing dataset, we estimate an industry model of calling, subscription, and pricing. Our counterfactual simulations, which incorporate operators' price responses to compute the new equilibrium, show that the regulation would lead to lower fixed fees, lower allowances, and lower overage prices chosen by the operators. The regulation and the pricing changes in response to it would benefit both the consumers and the operators, and overall the total surplus would increase by $307 million per month. An increase in mobile penetration explains the joint increase in firm profits and consumer surplus.