« Previous | Contents | Next »
Listen
SUMMARY OF MODELLING APPROACH
The main purpose of this report is to explain how the model works in some detail but it is helpful to provide the reader with an early fix on the underlying modelling approach before setting it out fully in Section 3. The modelling framework is similar in spirit to the Meen et al (2005) model for England. They delivered an affordability simulation model at standard regional level capable of providing simulated results over a forward period of 30 years via an accessible user interface constructed on Microsoft Excel. The underlying model represented an impressive and robust set of different, interacting models covering different aspects of the national and regional housing system.
The underlying idea for the English, and the Scottish, models is to distinguish between those elements that are endogenous (determined within the system of modules) and those elements determined outside of the system, exogenous variables, which can be the source of scenario testing (e.g. the impact of lower long term interest rates in the future). In practice, the different modules interact because, for instance, the outcomes of the household formation module directly shape housing demand and hence house price determination but they also interact with labour supply and provide the raw numbers that are divided into different housing tenure choices. At the same time, house price effects then influence migration decisions and these in turn affect labour market and future housing demand levels.
An important difference between the Scottish and English models is that the latter is based around standard Government Office Regions ( GORs) (e.g. the West Midlands or the South West), whereas Scotland is a unified GOR in its own right, and a different approach was required to arrive at cohesive functional geographies that operate at a sub-regional level. Section 3.3 sets out how the research team arrived at the final sub-national geography using a set of clear criteria to end up with eight local markets that sum to Scotland as a whole.
A further important difference between the English and Scottish versions concerns data availability and sample sizes for estimation purposes of the different modules. Although Scotland is much smaller than England, and this is magnified by the disaggregation strategy to local sub-regional market areas, and while this rules out the detailed estimation and use of certain data sources, Scotland does have specific data sources not available in England. More generally, however, the capacity to build the module estimations and then to link them sensibly is much more constrained in Scotland than it has been in the English case. The fact that the model solves and is as well behaved as it turns out to be is a testament to the research team's extended estimation and programming work.
What are the main high level strengths and limitations of the model? On the side of strengths, the model is an easy to use and comprehensive simulation model that allows users to simulate scenarios at the local and national level. It is best practice in terms of its estimating strategy (and has benefited from the considered input of the English model's research team leader, Geoff Meen), in the use of available data and diagnostic testing, and it is situated in the literature both conceptually and practically. Its main weaknesses are those of any simulation model that works on a long term time span. It is, by definition, less able to cope with short run unpredictable changes happening within the economic system. These unknown outcomes occur in time t +1 and then go on to have ripple effects changing parameter estimates in t +2 to t +n. Such models, therefore, need to be capable of re-estimation if we consider the impacts of unforeseen changes to be subsequently permanent as opposed to transitory and of less enduring importance. This also means that these models often do less well in terms of accurately capturing the amplitude and extent of cyclical, as opposed to trend, fluctuations. This is the essence and weakness of long term simulation models. This is why re-estimation with new and emerging data is so important if the model is to retain its relevancy and fit with the real economy.
« Previous | Contents | Next »