Skip to main content

Thomas Guerrero

Improving the predictive accuracy of Discrete Choice Models

Email: [email protected]

Supervisors

Project description

Endogeneity is a correlation between explanatory variables and the error term of an econometric model. As a result, the parameter estimates of these variables are inconsistent. It is an unavoidable problem in econometric modelling, with several causes. These include omitted attributes, measurement or specification errors, simultaneous determination and/or self-selection.

Discrete Choice Models (DCM) are fundamental in transport planning and policy formulation. Mode choice is the third stage of the classic transport model. It is typically modelled using DCM. These models are very much susceptible to endogeneity problems.

Endogenity problems have several causes. These include omitted attributes, measurement or specification errors, simultaneous determination and/or self-selection. All these causes are very common in transport modelling. If we don’t correct the DCM for the effect of endogeneity, estimates will be biased. This can produce unreliable forecasts and conclusion. So, if we base our analysis endogenous mode choice models, it may lead to wrong decision making.

Several methods are able to correct the problem of endogeneity in econometric models. But models that correct for endogeneity using these methods are scarce in the transport field. Still, there are some methods to correct endogeneity in DCMs.

In this research project, we will provide evidence to contribute to state of the art and correct this anomaly. We will develop three important goals:

  1. how to identify a valid correction for endogeneity (ie a strong instrument)
  2. how to validate the instrument
  3. the impact of endogenous modal choice models in urban transport modelling

Publications

Interests

  • endogeneity
  • discrete choice models
  • modelling transport
  • accident models

 

Programme information
Assessment and feedback
Student support