Use of mixed revealed-preference and stated-preference models with nonlinear effects in forecasting

Elisabetta Cherchi, Juan De Dios Ortúzar

Research output: Chapter in Book/Report/Conference proceedingConference contribution


Estimation of mixed revealed-preference (RP) and stated-preference (SP) data has become almost common practice. However, most attention has gone to estimation, with the correct use of these much-improved models in forecasting left in need of better understanding. In particular, their potentiality in accounting for nonlinear effect has not been fully explored, either in estimation or in prediction. The problem of using an RP-SP mixed model for accounting for nonlinear effects in the attributes in forecasting, especially when the nonlinearities have been specifically estimated only for the SP data, is addressed. By using a mixed RP-SP data set gathered especially for a modal choice context in Cagliari, Italy, several nested logit models with nonlinear systematic utility functions were estimated and used to provide empirical evidence on the effect of incorrect use of RP-SP specifications in forecasting under different policies. It was concluded that the partial data enrichment approach is well suited to treat nonlinearities and allows for consistent improvements in model estimation results. More important, nonlinearities might impose severe limitations in the applicability of these models in forecasting because of more restrictive microeconomic conditions. Failure to treat these problems correctly can lead to serious forecasting errors, even in the case of relatively mild policies.

Original languageEnglish (US)
Title of host publicationTravel Demand and Land Use 2006
PublisherNational Research Council
Number of pages8
ISBN (Print)0309099870, 9780309099875
StatePublished - 2006

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Mechanical Engineering


Dive into the research topics of 'Use of mixed revealed-preference and stated-preference models with nonlinear effects in forecasting'. Together they form a unique fingerprint.

Cite this