5. Sensitivity test for policy implication
A sensitivity test was conducted using the overall model shown
in Table 4 to identify the influence of travel distance (from an origin
to a chosen station) and travel cost (from a chosen station to a
destination) on the nearest station choice by holding other
independent variables at their mean. The resulting sensitivity
plot for all stations (Fig. 4a) indicates that the predicted
probabilities of choosing nearest station decrease as travel
distance increases for all five different travel fees, which are
travelling over one zone ($2.70), two zones ($4.00), three zones
($4.90), four zones ($5.80) and five zones ($7.10). Generally, the
closer the chosen station to the destination, the lower probability
of a chosen station is the nearest train station to the origin, except when travelling over four zones ($5.80). In the Perth metropolitan
area, only the Mandurah line extends over five zones. After closely
examining the travel patterns of respondents who travelled by
trains over four zones, we identified that 24% of them chose
stations, mostly Murdoch station on the Mandurah line, even
though they came from a location near other train lines.
Interestingly, about 82% of these travelled by bus feeder services
to the train station, which is a good example of commuter/workbased
transit service.
Fig. 4a also shows that the predicted probability of respondents
choosing a station that is the nearest station to their origin is over
80% if travel distance from an origin to the chosen station is less
than 800 m. However, when they have to travel over 10 km from an
origin to the chosen station, the estimated probability of choosing
the nearest station is still over 80% only for respondents travelling
over five zones. For respondents travelling over less than five
zones, the estimated probability dropped sharply, especially for
respondents who travelled on trains within one zone, the
estimated probability decreased to 39%. The station, which belongs
to the travelling-over-five-zone category, is Warnbro: a captive
station. While stations belonging to the travelling-within-onezone
category are non-captive stations. Lack of competition with surrounding stations has led to a bigger catchment area for
Warnbro station, leaving the train users with less travel options.
This demonstrates a certain level of transport disadvantage for the
train users.
The resulting sensitivity plot (Fig. 4b) for non-captive stations
calculated based on the model in Table 8 indicates the same trends
shown in Fig. 4. However, the probability of choosing the nearest
stations for non-captive stations decreased more quickly than the
captive station model. Generally speaking, for the same travel
distance, the more zones the commuters travel, the higher
probability to choose the nearest stations. A very interesting thing is the exception of travelling three zones ($4.9) and four zones
($5.8). The non-captive model shows that there was less than a 50%
chance for chosen stations to be the nearest station if travel
distance was 10 km. No travel over five zones ($7.10) was identified
for the transit trip involving in non-captive stations. This result
could be interpreted that expect for a distance minimisation
strategy, others such as cost and travel time minimisation
strategies and multi-trip purpose utility maximisation could play
important roles to the decision maker of this type trips. In addition,
more travel uncertainty, such as availability of parking, could be
involved in non-captive station choice than captive station choice.
A sensitivity test was also conducted using the logistic
regression model shown in Table 5 to identify the influence on
the nearest station choice for travel distance (from an origin to a
chosen station) and whether the station is further away from
origins and destinations. The resulting sensitivity plot for captive
stations, shown in Fig. 5, indicates that the predicted probabilities
of choosing the nearest station decrease as travel distance
increases, whether the station is further-away or not. Generally
speaking, the predicted probability of choosing the nearest station
is the higher for non-further-away stations than further-away
stations. According to Fig. 5, at 10 km travel distance, the predicted
probability of the nearest station choice is over 85% for nonfurther-away
stations. In comparison, it is less than 60% for furtheraway
stations.