Tri-reference-point hypothesis development for airport ground access behaviors

Abstract

Studies have applied single-reference-point or safety margin hypotheses to examine how advanced traveler information affects travel behaviors. However, these theories may fail to fully capture the trade-offs among origin departure time, airport access time, and terminal processing time in terms of airport ground access behaviors. In this study, we developed a tri-reference-point hypothesis and assumed that the rate of change of utility may change at the air passenger’s preferred (PAT), earliest acceptable (EAT), and latest acceptable (LAT) airport arrival times. With an empirical data set collected from 304 passengers at Taipei Songshan Airport, the study examined the tri-reference-point hypothesis by analyzing airport ground access mode choice behaviors with a pooled framework that combined revealed and stated preferences. Moreover, the study developed four alternative specifications for schedule delay variables, assuming that air passengers used different reference points to determine relative gains and losses of the expected airport arrival time. The specifications included selecting both EAT and LAT as the zero-utility points (an indifference-band specification) and either one of PAT, EAT, and LAT as the single zero-utility point. Regardless of which specification was employed for schedule delay variables, the tri-reference-point hypothesis was generally supported. In particular, a significant difference of the rate of change of utility around PAT, EAT, and LAT was identified in the analysis results. When managing increasing road travel times and increasingly congested terminals, air passengers were more willing to retime their origin departure time to an earlier time than to switch their ground access mode. The implications of the analysis results for airport ground access management are discussed in the study.

Appendix: Specifications and estimation results of the tri-reference-point model under earliest- and latest-acceptable-arrival-time-as-zero-utility assumptions

The utility change due to schedule delay variables under the EaZ assumption could be specified as follows:

$$ \Delta U_{qit} = \alpha_{SDE2} \left( {t^{e} - t_{it}^{a} } \right)SDE2_{it} + \alpha_{SDE1} \left( {t_{it}^{a} - t^{e} } \right)SDE1_{it} + \left( {\alpha_{SDE1} \left( {t^{p} - t^{e} } \right) + \alpha_{SDL1} \left( {t_{it}^{a} - t^{p} } \right)} \right)SDL1_{it} + \left( {\alpha_{SDE1} \left( {t^{p} - t^{e} } \right) + \alpha_{SDL1} \left( {t^{l} - t^{p} } \right) + \alpha_{SDL2} \left( {t_{it}^{a} - t^{l} } \right)} \right)SDL2_{it} $$

(10)

where $SDE1_{it}$, $SDE2_{it}$, $SDL1_{it}$, and $SDL2_{it}$ were four dummy variables, whose value was 1 if the expected airport arrival time fell in the extremely early, early, late, and extremely late segment, respectively, and 0 otherwise. Similarly, the utility change under the LaZ assumption could be specified as follows:

$$ \Delta U_{qit} = \left( {\alpha_{SDE2} \left( {t^{e} - t_{it}^{a} } \right) + \alpha_{SDE1} \left( {t^{p} - t^{e} } \right) + \alpha_{SDL1} \left( {t^{l} - t^{p} } \right)} \right)SDE2_{it} + \left( {\alpha_{SDE1} \left( {t^{p} - t_{it}^{a} } \right) + \alpha_{SDL1} \left( {t^{l} - t^{p} } \right)} \right)SDE1_{it} + \alpha_{SDL1} \left( {t^{l} - t_{it}^{a} } \right)SDL1_{it} + \alpha_{SDL2} \left( {t_{it}^{a} - t^{l} } \right)SDL2_{it} $$

(11)

The estimation results of these two models were summarized in Table 4.

Table 4 Tri-reference-point models: earliest and latest acceptable airport arrival time as zero utility