Publication Detail
Causal Models in Travel Behavior Research: Results of a Simulation Experiment
|
UCD-ITS-RR-93-18 Research Report |
Suggested Citation:
Pendyala, Ram M. and Ryuichi Kitamura (1993) Causal Models in Travel Behavior Research: Results of a Simulation Experiment. Institute of Transportation Studies, University of California, Davis, Research Report UCD-ITS-RR-93-18
Submitted for presentation at the 73rd Annual Meeting of the Transportation Research Board
Causal models are increasingly being used in travel behavior research as they offer greater insights into the underlying relationships driving behavioral phenomena. These models usually take the form of simultaneous equations systems in which several dependent and independent variables are related to one another in a set of equations. In travel behavior research, where causal theories of behavior are not established, various model specifications are estimated and compared in an attempt to identify causal relationships and formulate causal theories. This paper examines the validity of such an approach in light of the fact that any model specification is mathematically equivalent to several other model specifications. It is found that tests of competing causal hypotheses can not be conclusive, unless the tests are nested. The findings of the paper strongly point to the need to establish theories of travel behavior prior to estimation of models for planning and forecasting purposes.
Causal models are increasingly being used in travel behavior research as they offer greater insights into the underlying relationships driving behavioral phenomena. These models usually take the form of simultaneous equations systems in which several dependent and independent variables are related to one another in a set of equations. In travel behavior research, where causal theories of behavior are not established, various model specifications are estimated and compared in an attempt to identify causal relationships and formulate causal theories. This paper examines the validity of such an approach in light of the fact that any model specification is mathematically equivalent to several other model specifications. It is found that tests of competing causal hypotheses can not be conclusive, unless the tests are nested. The findings of the paper strongly point to the need to establish theories of travel behavior prior to estimation of models for planning and forecasting purposes.