The use of multiple data collection approaches to improve the validity of stated preference surveys

Abstract

Paper presented at the 22nd Annual Southern African Transport Conference 14 - 16 July 2003 "National issues affecting the movement of people and goods - strategic approaches", CSIR International Convention Centre, Pretoria, South Africa. ABSTRACT: A Stated Preference (SP) mode choice model was found to have a relatively low goodness-of-fit when estimated from semi-literate commuters. Some discrepancies were noted when the findings for the SP survey were compared with those from a qualitative (probing and observation) approach. The discrepancies concerned the effect on the validity of the data of two as opposed to three mode choice sets, three or five modal attributes, and the use of verbal or visual presentation formats. Detailed inspection of individual responses and in-depth ethnographic interviewing were able to reveal the probable causes of the poor goodness-of-fit, and so suggest ways in which SP surveys among less-literate respondents might be improved. Chief among the problems was that many of these respondents simply did not understand the concept of making hypothetical choices. The study clearly demonstrated the value of a multi-method approach to improve SP data validity.