Conclusions

In this chapter, the general modelling approach used in the dynamic spatial MSM has been described. The reason for this approach is to try to provide a better representation of the Leeds population through the more sophisticated updating/ageing of the individual attributes within a local context. Therefore the population is modelled as individuals within households that have a rich set of important attributes. Such households locate in wards that are the basic spatial units in the area model. Wherever possible, transition probabilities that drive the population changes have also been calculated at the ward level to reflect the local characteristics.

Six demographic processes are being modelled in this dynamic spatial MSM due to their importance in population evolution and planning reference functions. Each process has been developed into a separate module, but they can interact with each other. Various factors that drive the demographic changes have been considered accordingly in each process. Among them, age, sex and location have been selected as the foundational factors that need to be captured in the processes of all demographic changes. Extra factors are introduced according to the requirement of the individual demographic process. Detailed discussions have been provided on each 158

factor’s importance for the relevant processes. Such impact of the factors is captured in the probability calculation process. The probabilities are carefully calculated to reflect the chance of demographic changes for various groups of people with different characteristics. As described in previous sections, all six processes are probability driven and the simulation is based on the application of the Monte Carlo method.

The dynamic spatial MSM developed in this study provides the characteristics of studied population at the level of individuals through a truly dynamic ageing process. As described in Section 4.2.1, at each simulation interval of a year, the attributes of each individual change according to probabilities that reflect their demographic characteristics and local area characteristics. Such characteristics are captured in probabilities that are calculated using relevant demographic and spatial information. The Monte Carlo method is used to determine the probabilitys. This is the general method shared by all six demographic process simulations, namely: Ageing, Mortality, Fertility, Health Change, House formation and Migration.

At the aggregate level of Leeds, population age-sex structure diagrams have been generated using the initial results generated by the MSM. They demonstrate the patterns of the population changes year by year. As all changes are dynamically simulated each year and driven by multi criteria based probabilities, including local area factors, the overall results should present a much more robust representation of the studied population, compared to static models that move population forward through general reweighting procedures.

At the small area level of wards, it is found that characteristics of the local population changes differently in small areas. Local population in Cookridge clearly demonstrates a more serious problem of ageing than Headingley. Having used the LQ (Location Quotient) method to analyse the populations of a set of wards, it is confirmed that population evolution does vary geographically. However, it is also found that although the results from established suburban wards such as Cookridge seem to be more 159

plausible, the results produced in the current MSM are not satisfactory for wards that are largely impacted by student migration such as Headingley. Possible reasons have been discussed and issues found will be addressed in Chapters 5 and 6.

The results from individual processes have also been analysed. The spatial variation is found in small areas even in the same demographic process. Cookridge and Headingley have been chosen to represent two very different wards in Leeds and the simulation results from year 2030 are used in the analysis. In all 6 demographic processes, the two small areas demonstrated geographical variances as described in section 4.8. Due to the page limitation of the thesis, comparisons with the 2001 outputs are not discussed in the initial result analysis. Instead comparisons between the model results at the begging of the simulation (year 2001) and the end of the simulation (year 2030) are conducted using the revised model results in the following chapters 5, 6 and 7.

In short, the dynamic MSM models the individual changes using dynamic ageing technique. This means that each all simulated attributes are dynamically updated through the evolution of demographic changes and interactions of the demographic processes. At the aggregated level, the trend of the population changes can be traced over time. This can be useful for strategic decision making. At the disaggregated level, spatial variances have been found across different wards. The distinction also persists in all modelled demographic processes. Thus, the spatial MSM can provide valuable information for tactical decisions and location based studies and policies.

Migration is a more complex process because there are various reasons for migration. As there is interdependency between various demographic processes such as household formation, this process is more difficult to model. In the initial version of the dynamic spatial MSM, however, this process is only concerned with the flows within Leeds. Therefore the representation of the migration process has been flawed in this initial version of the dynamic spatial MSM. The lack of migrants that are moving

into Leeds, especially the younger population, may have caused the exaggeration of the ageing trend of Leeds. ONS projected that there are more migrants moving into Leeds than out. Also as the demographic processes interact with each other, a younger baseline population will have important impact on other processes. Consequently, the limitation in the migration module may also affect the population, births, deaths, health changes and household formations. Therefore, the issues introduced by the limitations in modelling of the migration process must be addressed. In Chapters 5 and 6, the further development of this process, attempting to address various issues, will be discussed in detail.