## Deterrence functions

If the method of
determining deterrence parameters was repeated using an even larger survey size,
then statistically significant results by trip purpose may be determined. This
would allow modelling of separate parameters based on each land use (journey purpose)
and transport mode, leading to an improvement in the quality of the derived
accessibility analysis.

An increased survey
size could be achieved by either immediately expanding the sample size of the NZHTS,
or waiting until the addition of successive years of data has sufficiently
increased the size of the NZHTS dataset.

Therefore, the
following recommendations are made:

·
The deterrence parameters that have
been developed for the transportation mode types for various OD types should be
revised when updated travel survey data becomes available.

·
The analysis of the deterrence parameters
should be extended to include other modes of transport such as taxi and train
once the sample size for these modes reaches an acceptable size.

Even so, the measurement of the deterrence parameters using the NZHTS assumes
‘measured’ behaviour reflects ‘desired’ behaviour. Strictly speaking this is
incorrect because measured behaviour already includes the spatial distribution
of the transport network and land use. Consequently measured behaviour is
already influenced by ‘need’, ‘accessibility’ and a number of other factors
specific to the individual undertaking the specific journey.

Although, in the
interest of pragmatism, deterrence parameters derived without reference to the
NZHTS would be more ideal, the current approach is considered appropriate. This
is an area of future study although in the interim, it is hoped that further
analysis and understanding of the NZHTS to account for the difference between
measured behaviour and desired behaviour will assist to determine and refine
the appropriate deterrence parameter values. Additionally it might be that
research regarding travel adaptive capacity assessment can be used to
supplement the NZHTS data (Krumdieck et al 2012).

Regardless, if the
NZHTS is used in its entirety or supplemented with ‘desired’ behaviour, the data should be tested using logistic decay
functions that are expected to produce better mathematical fitting models than
a negative exponential function (λ).