Consumer Demand Theory and Regional Air-Travel: an Integrated Economic & Econometric Approach (PhD thesis)
February 11, 2006 Leave a comment
The air-travel market is characterised by consumers making discrete choices for tickets in various ticket classes that are capacity constrained. These ticket classes are artificially created by the supplier, the airline, to discriminate between various types of consumers. The capacity constraint on the ticket class arises because of the limited number of seats that can be allocated on any one flight.
Standard economic models of discrete choice by consumers sees them choosing an alternative from a choice set that maximises their utility. The model that is proposed in this thesis adds a further dimension – the consideration of availability. Availability is included in the discrete choice by substituting utility with expected utility. The model that is developed includes consumers’ expectations of capacity and demand. The model also allows the consumer to manipulate availability by expending resources, much in the way suggested in manipulable rationing schemes.
It is shown that the consumer’s optimal solution for the expenditure of resources towards the manipulation of availability depends on the functional form of the components of the expected utility function. A set of comparative statics are provided to demonstrate the implications of a change in price, income, expected capacity and expected demand on the likelihood that a consumer chooses a particular alternative.
Critically, the comparative statics suggest that an increase in the consumer’s expectation of capacity for an alternative positively affects the likelihood that a consumer will choose that alternative. This particular comparative static result motivates the inclusion a variable that captures expected capacity in the later empirical analysis.
The empirical analysis that is reported in this research is built around a highly disaggregated dependent variable and a range of aggregated explanatory variables. A range of various modelling approaches is summarised. These include the use of an aggregated version of the multinomial logit model, a linear logit specification and a linear specification.
In examining capacity constraints, they not only have the potential to modify the behaviour of the consumer, but they also limit the observation of the actual underlying demand which biases estimates from standard econometric estimation methodologies. Given this, the Tobit censored regression is presented and suggested as a means to overcome the inconsistent estimates that would otherwise arise.
The thesis examines total sales and capacity constraint data for every ticket class on every flight between an Australian state capital, Sydney, and a major regional centre, Wagga Wagga, over the course of a financial year. This type of data also allowed the use of a more advanced method to improve the efficiency of the estimates. A multivariate Tobit model is specified, which uses a quasi-maximum likelihood technique to estimate a system of seven individual demand equations. The method that is outlined deconstructs the seven-dimensional likelihood integral into a likelihood function of many bivariate integrals. While this enables computation of the complex multidimensional integral, this computation remained costly and time-consuming.
The results of the estimation revealed that expected capacity was a determinant of demand. However, there was a range of specification issues that had to be confronted. For instance, the presence of heteroskedasticity was an issue in nearly all of the models. Conditional moment tests were used to verify the validity of the Tobit estimates. This was a particular problem because heteroskedasticity results in inconsistent Tobit estimates when those estimates are produced through the censored regression technique. Tobit models were necessarily re-estimated to correct for this diagnostic concern.
While the empirical models suffered from some diagnostic problems, they did provide an interesting insight into the behaviour of consumers in capacity-constrained markets. Expected capacity elasticities of demand were predominantly positive. Estimating demand equations for individual classes also presented an interesting picture of the preferences of consumers of different ticket classes. Given these supportive results, the research was shown to have ultimately met its three core objectives, which were:
1. to model the discrete choice behaviour of consumers when constraints reduce the likelihood of an option being available;
2. to review and to apply econometric methods that are applicable where data are not disaggregated sufficiently to allow standard discrete choice econometric methods; and
3. to provide estimates of a demand equation for a regional air route in New South Wales, Australia that both validates the first two objectives and provides an accessible empirically based resource for policy making.