Browsing by Author "Zhou, Hao"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Cooperation under Imperfect Monitoring and External Threat
    (2016-10-05) Zhou, Hao; Dudey, Marc P
    Cooperation among players is often a good deed to pursue. The famous "Prisoner’s Dilemma" in game theory has long been an example of showing how non-cooperation results because of conflict of interest among agents. My thesis investigates how factors like imperfect public monitoring, emergence of an external threat influences the cooperative behaviors among players in dynamic environments. Chapter 1 investigates bidder collusion in repeated procurement auctions without communication or side payments, focusing on the case of bidders having identical costs under imperfect public monitoring where only winners, not bids, are publicly observed. It presents a simple bid rotation scheme, in which bidders take turns entering bids equal to the auction’s reserve price. This behavior is supported by the threat of a bidding war, i.e. if an auction is won by the wrong bidder, all bidders will enter bids equal to their common production cost in all future auctions. Bid rotation maximizes the bidders’ joint profits and is a perfect public equilibrium if and only if the discount factor is greater than or equal to a critical value, call it delta. This paper presents two main results. First, except for a measure zero set of discount factors, joint profit maximization cannot be achieved for discount factors below delta by any profile of bidding strategies. Second, in the case of two bidders, there is no profile of bidding strategies that achieves joint profit maximization for discount factors less than or equal to delta. Chapter 2 investigates a situation where firms selling different products refer mis- allocated customers to one another. Under monitoring over each other’s sales in every period, we analyze firm referrals in an infinitely repeated game with by looking at a class of “k + 1 punishment schemes”, in which players “forgive” the first k bad signals, and “punish” each other forever after the k + 1’s bad signal. We characterize the unique optimal k in this class of schemes. Chapter 3 is an empirical paper that models the impacts of telemarketing calls for selling bank long-term deposits. We use a dataset from a Portuguese retail bank from 2008 to 2013. This dataset contains features related to the calls and customers. We model a binary response of the outcome of the telemarketing call (yes or no) using those features. In second part of the paper, we propose methods to model price elasticity of demand (PED), which measures sensitivity of the long-term deposits resulting from changing interest rates. In estimating the PEDs, propensity-score-matching is used to adjust for potential group differentiation. Chapter 4 studies alliance behavior under external threat. When an alliance faces danger of appropriation from an external enemy, it is optimum for its members to jointly invest in their defense. For members to behave collusively in the subgame perfect Nash equilibrium (SPNE), each of them has to be allocated with some minimum share of that resource. Under proportion-to-share rule of cost contribution and profit earning, this paper looks at how that minimum share requirement changes after the emergence of the external threat. We find that two factors, alliance size and cost factor contribute to the stability of the alliance in opposite directions. Further, force from more costly investment outweighs the force from increasing alliance size, which makes the alliance easier to maintain.
  • Thumbnail Image
    Item
    Stochastic simulation for viral diseases: Dengue and avian influenza
    (2008) Zhou, Hao; Deem, Michael W.
    Our immune system protects us against viral invasion. Nevertheless, variability helps viruses to escape from immune suppression. Viruses may vary either by existing in multiple subtype forms or by mutating at a non-negligible rate. Effective design of vaccines for viral diseases requires some estimation of the escape mechanisms as well as the variability of the virus. However, prediction of viral escape is non-trivial due to the complexity of the virus system. I propose to use stochastic simulation to model viral diseases at the sequence-level. I focus on dengue virus and avian influenza. Dengue virus, having four closely related serotypes, represents a prime example of where vaccine development has been delayed because of incomplete understanding of the immune response to multiple components. Here I extend the generalized NK model theory to study dengue virus dynamics. I elucidate the mechanism for two puzzling phenomena: original antigenic sin and immunodominance. I also suggest new polytopic vaccination strategies to protect against the four serotypes of dengue virus. Previously generalized NK model was applied to B cell immunity. I extend the model to apply to T cell immunity, which is important in the control of dengue virus. Influenza is a highly mutating virus. Influenza pandemic occurs periodically. The H5N1 bird virus is regarded as one candidate pandemic-causing strain. Various countries around the world have started to create stockpiles of H5N1 avian influenza vaccines. However, since the avian influenza is mutating and multiple virus introduction is a possibility, how many and which strains should be stockpiled? I analyze the influenza database and simulate the avian influenza virus evolution to obtain optimized strains combinations. I also suggest population at risk (PaR) as a metric for infectious disease risk management. For both dengue and influenza, simulation results can not only help us to assess new vaccines and provide strategies for specific diseases, but also provide us with insights to other diseases, such as HIV and cancer.