To reduce traffic along a heavily congested Texas toll road, researchers are using stated preference surveys and discrete choice models to understand how motorists will respond to alternative time-of-day pricing policies.
Quantitative Tools: Discrete Choice Modelling
The Choice Modelling and Valuation Group has made major innovations and extensions to best practice in the area of discrete choice modelling, as a means by which to understand and predict choice behaviour.
Where appropriate we develop models using information from choices that people are observed to make, referred to as revealed preference (RP) information. In cases where there are limitations in the information provided by observed choices, for example when predicting demand for a new product or service, we collect stated preference choice information (SP data).
A number of procedures for eliciting stated responses exist, but current practice focuses on the use of discrete choice experiments (DCE), which involve the presentation of hypothetical choice situations to respondents in surveys. In SP DCE each alternative is described by its relevant attributes, for example, the quality of the service, the cost of the service, future characteristics, etc. Each of the attributes in the experiment is also described by a number of levels, e.g. low cost versus high cost. The attribute levels are combined using principles of experimental design to define different packages of goods or services which individuals then compare in surveys.
The outputs from discrete choice models can be used to improve understanding of the drivers of people’s choices, including:
- estimates of the relative importance of different attributes for a specific product or service, for different population groups
- estimates of the trade-offs or marginal rates of substitution that people are willing to make between attributes, providing indirect measurements of willingness to pay.
The outputs from discrete choice models can also be used to develop predictive models to gain insight into how people's choices may change under differing circumstances. Using these models is particularly useful for policymakers to demonstrate the likely impacts of a policy. The models can provide estimates of changes in demand for services, as well as insight into how a policy may impact different groups within society. The models can also quantify how individual attributes influence demand, thereby providing estimates of elasticities, as well as providing estimates of consumer surplus, i.e. monetary valuation of the benefits obtained from different services.
An important benefit of the rigorous statistical procedures we employ is that information can be given on the accuracy of all the outputs of the model, as well as indicating whether specific aspects of the choice are truly significant in influencing behaviour.
RAND Europe was one of the first companies to employ SP DCE in the transport sector and continues to conduct research to improve SP methods. RAND Europe pioneered procedures to combine RP and SP data to exploit the strengths of each of the data types to best advantage. We also offer expertise in RP and SP survey design, based on insights gathered from our extensive practice in RP and SP modelling.