With the aim of increasing realism for respondents in Stated Choice (SC) studies, analysts increasingly include a reference alternative in their surveys, with the attributes of the remaining, purely hypothetical options, being pivoted around those for the reference alternative. However, a problem with this approach is that the presence of the reference alternative may lead to significantly higher choice rates for the reference alternative, which can, in the worst case, express itself as non-trading behaviour. Such inertia is common, especially in the case of surveys administered to commuters whose reference alternative will be the result of years of fine-tuning their journey. Problems may also arise when the reference alternative is an untolled option while a toll is incurred for the remaining options. In this paper, we make use of data collected recently in the United Kingdom using a survey that includes a reference alternative in a route choice context, and where the non-reference alternatives additionally occasionally incur a road toll. The presence of the reference alternative was judged to be beneficial in that respondents were able to relate to the choice sets in what was a highly complex design, involving up to six attributes and up to five alternatives, where the complexity varied across respondents. However, at the same time, high rates of inertia where observed, with, in some segments, rates of choice for the first alternative of up to three quarters. This situation had been anticipated at the design stage, and any respondent found to select his or her current option was then asked to additionally make a choice between the remaining hypothetical alternatives. Our analysis shows that augmenting the data by including these secondary choices leads to significant gains in robustness for the marginal utility coefficients, where, in the primary choices, a large share of the behaviour was explained solely by inertia, i.e. captured in the alternative specific constant. In addition to explaining the benefit of including information from the secondary choices, the paper also includes a detailed comparison of the results obtained between the primary and secondary choices, including scale differences. Finally, the paper discusses how the nature of the design, in terms of numbers of alternatives, attributes and choice sets, affects the rates of trading and hence reliance on the secondary choices and also shows the relative quality of results estimated on the primary and secondary choices across the different designs.
