What is the behavioural response impact difference of a push-pull policy initiative in contrast to a standalone push or pull initiative?

Source abstract

It is common practice to utilise elasticity estimates associated with a specific policy initiative such as a public transport fare or service frequency change, or the introduction of a revised road user charge for passenger cars. In reality, we often observe more than one initiative being acted on in a push-pull policy setting. It is questionable that the analyst can take a push response and a pull response based on standalone mean probability elasticity estimates and simply aggregate the net impact as the overall behavioural effect obtained after accounting for the directional effects of each of the push and pull initiatives assessed together. Not only is the mean effect likely to be different to the effect jointly assessed, but that there are likely to be noticeable differences associated with the socioeconomic and trip making activity of individuals and households. Drawing on data collected in Australia in May 2025 that investigated whether an initiative has a positive impact, a negative impact, or no impact on an individual’s travel behaviour, we show that when two policy initiatives are combined in a push-pull pair, the mean probability elasticity response is very different for the influencing socioeconomic and trip activity explanatory variables (identified in a multinomial logit model), than when the pull and push measures are analysed separated (using two MNL models).