Abstract:
It has always been di cult to analyse the premium from single ownership to multiple
ownership because usually we cannot control the same land tenure system to analyse the
in
uence of number of land titles on property price. In New Zealand, there is a special
land system called cross-lease. It can be compared with freehold because they all have
permanent land types; freehold is a single ownership and cross-lease is representative of
multiple ownership. In this way, we can more accurately analyse the premium from single
ownership to multiple ownership. In addition, this project has another goal. First, it is
well known that unless land use is allocated to completely independent plots, it is necessary
to manage public property (such as infrastructure) and allocate related external costs. In
this project, I compare the price di erence between a house with unit title and a house with
cross-lease rights to get the external cost of these management. Therefore, the second aim
is to analyse the price di erence caused by this regulatory cost while controlling number
of ownership. We use linear models and tree models to analyse them. Since the tree model
is relatively complex and cannot be directly analysed, we use the SHAP technique. This
technique can be used for any machine learning model and is a very versatile method.
Our analysis results of all models showed that the unit titles with designated supervision
schemes usually have higher prices, while the linear model showed that when the number
of ownership is small, the supervision cost is lower, and with ownership as the number
increases, the cost of supervision also increase. Finally, all models demonstrated that
single ownership has a higher price compared to multiple ownership. The linear model
clearly pointed out that as the number of ownership increases on permanent land, the
price will continue to decrease. In the tree model explained by SHAP, there is no clearer
downward trend. Eventually, through the evaluation of correlation analysis, I a rmed
the rationality of linear model analysis, and analysed related reasons that may lead to
imprecise SHAP analysis, as well as some conjectures about SHAP defects.