Before moving on, the choice of split point in Model (1.1) deserves further discussion.
As mentioned, the date 31/12/2000 was chosen because roughly 50% of the
data occur either side of this date. However, Model (1.1) could easily be fitted using
alternative dates for the split point and/or data-driven methods for choosing an
“optimal” split point. Another possible date for a split point would be a date that
represents when the impact of Government regulation is believed to have taken
affect. An optimal split point could be found by considering a sequence of potential
dates, and the optimal split point chosen to be the date which corresponds to Model
(1.1) minimising a given model selection criterion.