A GIS-based examination of residential dwelling figures in Newcastle Upon Tyne : comparison of the 1991 census and the local authority housing data
The study compares the total number of dwellings in Newcastle-upon-Tyne according to the 1991 Census and the local authority Housing Department. This is to assess whether the explanation of the differences at the local scale, can contribute to the understanding of Census underenumeration at the national scale. The significance of dwelling figures is established by reference to Census underenumeration, the Estimating with Confidence (EwC) project, housing need and government finance. The study also draws on literature about GIS in local government in order to highlight the need for the local analysis of dwelling figures for research and policy purposes, and to demonstrate the benefits of integrating Census and local authority data. The study describes the processes of data collection, integration and interpretation from both the Census and local authority sources. Total residential dwellings, occupied council stock and total vacant dwellings from both sources are compared and the differences are standardised. The standardised differences are then mapped (using z- scores) at the District scale and at Housing Area, Housing Neighbourhood, Ward and ED resolutions. Areas with large differences above or below the expected differences are highlighted and their characteristics are noted. The observed differences are then correlated with some of the characteristic of the areas in the form of dwelling structure, tenure, imputed residents and accommodation not used as main residence. These characteristics are then compared with those of areas with EwC non-response adjustment figures. The correlation analysis is carried out at two scales with the same resolution. This includes all the EDs in the City, then focuses on the EDs within Benwell and Scotswood Wards. This is to highlight the significant associations, specific to these Wards. Benwell and Scotswood are selected because of their patterns of vacancy differences, which are found to be in contrast to the city-wide pattern. The Census excess of vacancies in these Wards are also found to be greater than those in other Wards and greater (more than 1.4 standard deviations) than the city-wide expected average difference. The correlation analysis at the City scale finds that EDs with higher Census vacancy counts than the local authority data, share similar characteristics with EDs containing high EwC non-response adjustment figures. These are EDs with greater number of flats in residential buildings and local authority rented dwellings. In Benwell and Scotswood this pattern changes to greater number of flats in commercial buildings, converted flats and privately rented dwellings. The last stage of the study explains the differences in the selected variables using the individual property records (ED profiles). The study finds that at District (City) scale, the Census counts of total residential dwellings and occupied council stock are slightly higher than those from the local authority data. The Census vacancy figures however, are significantly less than local authority figures, mainly due to definitional differences and data collection methods used. The opposite pattern of higher Census vacancy figures is observed in Benwell and Scotswood. The reason for this is found to be due to vacant dwellings awaiting demolition, which were included in the 1991 Census vacancy figures but not in the local authority datasets. An example of how different definitions and data collection methods caused the observed differences. The study illustrates that the comparison of local authority and Census data can highlight areas with large differences (in vacant, occupied council stock or total residential dwelling counts) through data standardisation. The characteristics of these areas are found to be similar to those of areas reported as difficult to enumerate, in the national studies of Census underenumeration. The study also illustrates that these differences can be associated with Census non-response an...