The Design of a Large Scale Airline Network

Abstract

Airlines invest a lot of money before opening new pax transportation services, for this reason, airlines have to analyze if their profits will overcome the amount of money they have to invest to open new services. The design and analysis of the feasibility of airlines networks can be done by using the models developed in this thesis. It is possible to identify routes that are opportunities to open services, based on airline operating costs, passenger demand, aircraft and airport capacities and constraints. The optimization model based on the maximization of the net present value (NPV) allows airlines to make sure that a set of routes and links forming new networks represent an investment. The different airline business models, strategies, advantages and disadvantages were identified from the literature review. In general, four generic passenger business models are commonly recognized: full service carriers (FSC’s), low cost carriers (LCC’s), charter airlines and regional airlines. In reality, each airline has its own business model. As a main conclusion, it is also possible to say that the low-cost long-haul airline business is not competitive against full-service long haul and charter airlines businesses. The fare is the most important parameter or variable. Airlines can use it as a tool to get into the market, and increase passenger flow. Other parameters that airlines must consider in a model or methodology that determine what routes to operate are the target market, network structure, size and power of the route network, optimum number of passengers to attend, aircraft utilization, aircraft physical characteristics, size of the fleet, aircraft route load factor, seat capacity, frequency, turnaround times and airport capacities and infrastructure. Two airline operating cost models were developed based on econometric and engineering approaches. Both models were compared with real data. They were found to be highly accurate to calculate airline and aircraft operating costs. First, the AGE model (Chapter 3) is the most suitable model for the purpose of this research because it differentiates between aircraft operating costs. The AGE model can calculate route aircraft operating costs per flight per aircraft. It is based on load and range and it is highly accurate to calculate the total jet fuel volume needed to fly a certain route link by any aircraft type. The AGE model was used to generate operating costs data in the studied cases of Chapter 7. The second model is a translog function (AOC model), also developed in Chapter 3. It most important advantage is the possibility to calculate different operating costs per route per pax. It allows studying each operating cost effect on fares. The AOC model represents some advantages analyzing airlines market behaviour such as airlines business models competition. The disadvantage is that it does not consider that different aircraft types have different costs. A model that determines the lowest possible average route fare was proposed in Chapter 4. The model purpose is to identify routes that could be an opportunity to open services. Calculating a route fare is an important managerial tool. Different parameters have been identified as possible airfare determinants based on the literature (Table 4.2). Two mathematical models were developed in Chapter 4. Both models determine airline fares: the fare estimation model (FEM) model and the CFEM model methodology. The FEM model is a union between the AOC model (Chapter 3) and airport fees, social, economic and competitive factors. The statistic tests (ANOVA, t test and F test) confirm that the model can calculate fares and its variables have an influence on airline fares. In the case of a new airline, the FEM model can assume its operating cost factor by comparing with a similar airline business model because it is unknown. Although, the FEM model calculates airline routes fares highly accurate. The objective is to calculate the most convenient fare to identify routes where an airline can open new services. The FEM model can calculate the average route fare but not the range in between airlines route fares are expected to be. This is important, under competition airlines need to know how cheap the other airlines fares can be. Airlines can identify routes to open new service by knowing the other airlines possible min and max route fares. The CFEM model has been developed to calculate these ranges (Chapter 4). The CFEM model calculates the competitive fare using coefficients calibrated for the FSC-LCC market. In the case of an existing airline, this value needs to be compared with the FEM model average fare calculation. The comparison tells us if the CFEM model average fare calculation is possible. In the case of a new airline, the FEM model can calculate the operating cost factor needed to achieve the CFEM average fare value. Thus, both models complement each other. The CFEM methodology determines what routes represent an opportunity to open services by an airline based on the calculation of the most competitive fare. Nevertheless, it is not enough to find routes where the pax demand is high or good enough to open services. The forecast of the air pax demand is important for economic decisions of network planning, fleet assignment, new routes and investment. For those routes, it is very important to know if the induced demand is higher than the demand that is already transported by the other airlines operating the route before making the decisions of opening new services. In reality, the induced demand cannot be known. The induced demand estimation model (PEM) simulates the pax flow behaviour (Chapter 5) by describing the distribution of the pax flow behaviour using approximately 18,000 routes data points. It was found that the log-normal distribution function (Equation 5.6) is the function that describes the US Domestic market behaviour better. The database was divided into five different airports types depending on the total number of pax transported per day. The results confirmed that the pax flow can be simulated by the log-normal distribution function. The routes that represent an opportunity to open new services are selected using the next criteria. Routes where the actual number of pax flow is higher than the PEM model calculations do not represent an opportunity to open services. In these routes, it is believed that most of the demand is already being attended. On the other hand, routes where the actual number of pax flow is smaller than the PEM model calculations represent an opportunity to open services. In these routes, it is believed that part of the demand has not been attended. Even when one route has been selected by the CFEM and PEM models, it does not mean that they represent an opportunity to open services. The routes selection of an airline network have more constraints than just finding routes that are very expensive and have a high number of induced passenger demand. Aircraft and airport capacities, characteristics, and limitations are main constraints to consider when selecting routes that represent an opportunity to open services. Airlines have to make sure that their set of routes and the links forming their networks represent an investment. In this thesis, the optimization model integrates aircraft performances and airport capacities with two financial methods: net present value (NPV) and internal rate of return (IRR). The return on investment (ROI) is also used to compare which aircraft and network generates more benefits for their required investment. The ROI is not a constraint, and it is not part of the optimization model. The importance of calculating the ROI is analyzing the efficiency between aircraft after optimization. The maximization of an airline network net present value (NPV) requires forecasting the pax demand for future years. The NPV optimization model is presented in Chapter 6. In this thesis, a Grey model (GM) modified version has been used to forecast routes pax flow for the long-term. The GM model was selected because it has the capacity to forecast data that have unknown parameters and it requires few data to approximate the behaviour of unknown systems. In the model, frequency is affected by the pax perception of travelling. This value represents the pax willingness for travelling the link. The number of frequencies increases on a link if the perception value increases. It decreases if the perception value decreases. The number of frequencies is also related to the aircraft operating costs. Reason why, the frequency calculated minimizes operating costs per route link. The model can analyze two different airlines network (NTW) cases for each aircraft type. On one side, the model allows studying a network without links with negative profits. On the other side, allowing aircraft to operate links with negative profits is relevant because an airline can earn more money. The generation of both types of networks is important to compare if different sets of routes with a different set of links results into higher net present values. A route generation algorithm assigns a route number and a link number to each possible link previously selected by the CFEM and PEM model. The objective is to determine the aircraft path. The algorithm considers the elimination of those routes that do not break the flying path only when a route NPV is negative. This algorithm achieves two things. One, it increases the number of pax transported by the airline. All possible routes are considered for open services when the NPV of a route is positive. Two, the route will eliminate routes with negative profits to find the links that make it profitable, whilst flying the maximum possible number of connections. The generation algorithm does not generate all possible routes between the city/airports in an airline network by purpose. Then, it is possible that another combination, with less routes and links, has a higher NPV than the route generator best combination, but transporting less number of passengers. The objective of this thesis is to find routes that represent a good opportunity to open services and design airlines networks by maximizing the network NPV and the total number of passengers transported in the network. It means to provide service to most of the links selected by the CFEM and PEM model, if the NPV still suggesting a good investment. The model assigns the optimum aircraft type to each route in an airline network. The model also applied short-haul and long-haul strategies. It allows the model simulating short-haul and long-haul scenarios. The results indicated that direct flights carrying cargo are a better business than connecting passengers through hubs. It means that it is very difficult to create an airline company that operates long-haul low-cost networks. The long-haul low-cost model connecting LCC’s companies in different regions through hubs allows increasing airlines networks net present values. The cases studied in Chapter 7 have shown that carrying cargo is what allows long-haul airlines to earn more money, but this business is already operated by FSC’s and cargo companies. Finally, the cases studied in Chapter 7 show that there are not an ideal aircraft. It is possible to find the optimum aircraft for each network.