The 12th and 13th of July are public holidays in Northern Ireland marking the Battle of the Boyne between the armies of James II and William of Orange in 1690. I spent most of the holiday trying to produce a map of population change in Belfast between 1971 and 2001.
Over the 20th century the population of urban areas in the UK first increased, in a process termed urbanisation, then in the early 1960s stagnated, and then reversed as urban areas started to lose population to their more rural hinterlands, a process termed counter-urbanisation. Some geographers have described modern cities as having a post-modern form meaning that the form of cities has become diverse, fragmented and unstable, rather than being organised around a defined city-centre. Other geographers maintain that at a local level residential neighbourhoods are still strongly stratified along socio-economic lines with important implications for resident’s life-chances. Looking at these ideas critically needs data on changes in the structure of cities over a period of at least 30 years and hence I have recently been working with Northern Ireland census data from 1971 to 2001.
From a professional viewpoint one of the main problems in studying processes of population change is that census data, which is the only reliable data we have for the population of small areas, is not available for a consistently defined set of areas over time. The definitions of small areas are generally different between each census. Although methods have been developed to produce population estimates for a consistent set of geographical area, such techniques are not really useful for producing population estimates for local neighbourhoods. In Northern Ireland, it is easier to study local population change, however, because the statistical office releases census data for a series of grid squares, either a square kilometre, or in urban areas, an area of 100 square metres. Every address recorded in the Census in Northern Ireland is identified by its grid reference and the original household data is used to describe the socio-economic characteristics of the residents of each grid square. This is likely to be much more reliable than trying to apportion data describing the characteristics of one set of areas to a second set of areas, say according to the number of postcodes they share, as is done in England and Wales.
Using the Northern Ireland grid data let's quickly look at population change in the central part of Belfast. The figure below is a histogram showing the change in the number of people resident in each grid square in central Belfast between 1971 and 1991 (in total, there are around 1,800 grid squares). Some grid squares gained population but a majority saw a decline in population. The magnitude of the change in population in each grid square is typically not very large with a median change in population of around 35 people, but in total this adds up to a significant loss in the population of the city.
In order to look at the geography of population change we could produce a map of the change in population for each grid square. The level of detail in this map would make it difficult to understand, however. An alternative is to produce a smoothed map of the change in population using a technique called kriging. Kriging first describes the spatial variation in population change between 1971 and 2001 using a statistical model and then uses the model to predict how the extent of population change varies over space. We don’t have to predict population change at the same scale at which we fitted the model (100 metre grid squares) but we can use any set of points we wish. This idea is common in many areas of statistics: the raw data is noisy with many small and not particularly interesting changes, for example changes in population due to new roads, demolition of old houses, the building of new ones etc. and we want to use the model to remove the impact of those small changes in order to see more clearly what the underlying structure of population change looks like.
The figure below shows the pattern of population change from 1971 to 1991 across the central part of Belfast predicted by the model. This shows clearly that areas in the west of the city have both experienced significant population decline (the blue areas) with another area of population decline to the south of the old docks, an area now occupied by the Odyssey Arena.
Population loss from the city has not been random but has followed a definite pattern. In the UK at least, cities may have become more fragmented but local neighbourhoods are still strongly distinguished by their level socio-economic characteristics. In this sense at least, Belfast is similar to other cities in the UK.