This thesis is composed of two parts, each of which reflects our attempt to describe order flow determinants in a bilateral and multilateral trading environment, respectively.
In Part I of this research, we investigate noncooperative bilateral sequential bargaining games in which the value of the asset changes stochastically according to a sequence of perfectly observable time-varying random variables. We attempt to model scientific speculations of the game participants that lead to varied length of bargaining durations. Previous studies, which have focused on the analyses of incomplete information games in interpreting bargaining delays, have shown that such delays are attributed to information asymmetry on asset values among players that results in differences in players' personal valuation of the asset. However, following the viewpoint of the Efficient Market Hypothesis, we assume in our models that there is no uneven assimilation of information of vital importance that affects the asset value once the players are at a negotiating table. Hence, one of the important features of the investigated models is that both players observe identical information regarding the future asset value, and that there is no uncertainty regarding one's opponent's preferences during the bargaining process. Despite the assumption of complete information, we argue that a delay before an agreement under certain conditions is an inevitable consequence of the stochastic component in this model.
We give game theoretic specifications for two types of bargaining games, which we call the Basic game and the Alternative game. The two games differ from each other in their timing of information arrivals with respect to players' actions. We characterize their subgame perfect equilibria that follow our particular behavioral assumptions. Characteristics of the equilibrium outcomes of the two games are compared. We direct special attention to the study of the analytical results in comparison with those of Rubinstein (1982), Osborne and Rubinstein (1990), and Merlo and Wilson (1995). We then give statistical specifications for two types of stochastic bargaining simulations, which are the Autoregressive Binomial Model and the Generalized Wiener Process Model. Comparative statics of several variables and bargaining durations are investigated thoroughly through numerous simulation runs. Subsequently, through our research we clarify the importance of integrating stochastic concepts into the bargaining theory and its applications in search of alternative explanations for various bargaining durations.
In Part II of this research, we provide a set of experimental results in our study of order flow determinants in experimental financial markets with asymmetrically informed human subjects. The markets are organized as computerized double auctions accommodated with an order book that contains a complete set of current limit and market orders and that can be inspected by every market participant at any time during each trading period. Our empirical analysis focuses on the series of actions taken by the subjects that include quote revisions, limit order arrivals, and trades. At first, we report thorough descriptive statistics on the extracted data sets, where we do not assume any particular theory of the market microstructure. Then we show serial dependencies of order flow on the previous event type, the state of the order book, the size of bid-ask spread, and the time intervals. In so doing, we ascertain the significance of the impact of information carried in the order book.
