Transportation systems such as mail, freight, passenger and even telecommunication systems most often employ a hub and spoke network structure since correctly designed they give a strong balance between high service quality and low costs resulting in an economically competitive operation. In addition, consumers are increasingly demanding fast and reliable transportation services, with services such as next day deliveries and fast business and pleasure trips becoming highly sought after. This makes finding an efficient design of a hub and spoke network of the utmost importance for any competing transportation company. However real life situations are complicated, dynamic and often require responses to many different fixed and random events. Therefore modeling the question of what is an optimal hub and spoke network structure and finding an optimal solution is very difficult. Due to this, many researchers and practitioners alike make several assumptions and simplifications on the behavior of such systems to allow mathematical models to be formulated and solved optimally or near optimally within a practical timeframe. Some assumptions and simplifications can however result in practically poor network design solutions being found. This thesis contributes to the research of hub and spoke networks by introducing new stochastic models and fast solution algorithms to help bridge the gap between theoretical solutions and designs that are useful in practice. Three main contributions are made in the thesis. First, in Chapter 2, a new formulation and solution algorithms are proposed to find exact solutions to a stochastic p-hub center problem. The stochastic p-hub center problem is about finding a network structure, where travel times on links are stochastic, which minimizes the longest path in the network to give fast delivery guarantees which will hold for some given probability. Second, in Chapter 3, the stochastic p-hub center problem is looked at using a new methodological approach which gives more realistic solutions to the network structures when applied to real life situations. In addition a new service model is proposed where volume of flow is also accounted for when considering the stochastic nature of travel times on links. Third, in Chapter 4, stochastic volume is considered to account for capacity constraints at hubs and, de facto, reduce the costs embedded in excessive hub volumes. Numerical experiments and results are conducted and reported for all models in all chapters which demonstrate the efficiency of the new proposed approaches.

In this dissertation, I investigate hub carriers' competitive advantage in directional markets within their hub and spoke networks. In the first chapter, I examine whether the competitive advantage of hub carriers in attracting hub-to-spoke passengers relative to spoke-to-hub passengers affects rivals' entry decisions in a symmetric way. The hub carrier advantage in attracting passengers at its concentrated hub airport creates an environment in which variation in the composition of demand in hub-to-spoke markets affects entry in a profoundly different way than demand variation in spoke-to-hub markets. In the second chapter, I examine hub carrier fares and price-cost margins in hub and spoke airport pairs. Exploiting variation across airport-pairs, I find that an increase in the proportion of business travelers in hub-to-spoke markets increases fares in these markets, while an increase in the proportion of business travelers in spoke-to-hub markets decreases fares. This result is consistent with the structural asymmetries found in the first chapter. However, the source of these concentration advantages remains ambiguous. These advantages could be due to cost benefits, demand effects, or market power. Exploiting the variation between hub-to-spoke and spoke-to-hub markets within airport pairs isolates the market power effect on fares. I find that difference in hub carrier airport shares explains most of the variation in its hub-to-spoke and spoke-to-hub price-cost margins. Unobserved quality and cost heterogeneity do not bias the result. In the final chapter we look at the relationship between market power and price discrimination. In the presence of price discrimination, at least one price does not equal marginal cost. Therefore, if price discrimination exists, there must be market power. While this logic is sound, it has led many policymakers to believe that price discrimination and market power are positively correlated. We present a model where measured price-discrimination can be low while market power is high and price discrimination can be high while market power is low, thus demonstrating that there is no theoretical connection between the strength of price discrimination and that of market power. We then present new evidence that price discrimination is negatively correlated with market power in the US airlines industry. / text