Hybrid discrete choice (HDC) modeling requires indicators to allow for the identification of latent variables. An indicator usually expresses the level of agreement of a respondent with a given statement, generally based on a Likert scale response. Literature exhibits remarkable variations regarding indicators’ complexity, expressed by the number of indicators for each latent variable, the type of scale, and granularity. Dealing with different levels of complexity implies that respondents require different cognitive efforts when choosing a relevant Likert point. Further, as a Likert item is undoubtedly a set of ordered categories, modelers face the challenge of estimating a large number of threshold parameters resulting from greater complexity when using ordered models for measurement equations. This paper studies the influence of indicators’ complexity on the estimation of HDC models based on an experiment that systematically varies the number of indicators, the granularity and the type of scale. We specified a proper parameterization of the scale factor of the measurement component for capturing potential effects of complexity in measurement equations. Findings revealed that granularity and its quadratic effect, as well as the interaction between granularity and number of indicators, affect the error variance of the measurement component and have a substantial impact on the goodness-of-fit of the discrete choice sub-model. Modeling results also showed that using odd-numbered scales, widely used in transportation choice studies, contribute to a lower error variance of the measurement component.
