Abstract
Recently, a literature has emerged using empirical techniques to study the evolution of international cities over many centuries; however, few studies examine long-run change within cities. Conventional models and concepts are not always appropriate and data issues make long-run neighbourhood analysis particularly problematic. This paper addresses some of these points. First, it discusses why the analysis of long-run urban change is important for modern urban policy and considers the most important concepts. Second, it constructs a novel data set at the micro level, which allows consistent comparisons of London neighbourhoods in 1881 and 2001. Third, the paper models some of the key factors that affected long-run change, including the role of housing. There is evidence that the relative social positions of local urban areas persist over time but, nevertheless, at fine spatial scales, local areas still exhibit change, arising from aggregate population dynamics, from advances in technology, and also from the effects of shocks, such as wars. In general, where small areas are considered, long-run changes are likely to be greater, because individuals are more mobile over short than long distances. Finally, the paper considers the implications for policy.
Notes
1. See www.rbkc.gov.uk/vmhistory/general/default.asp for more details.
2. This also allows us to provide more information on whether the samples are truly representative of the area statuses.
3. Robustness checks included setting the dependent variable to unity only if the MSOA moves more than one quartile, on the grounds that changes to an adjacent quartile might simply reflect the sampling. This alternative did not affect the results qualitatively and, so, are not reported here.
4. Note that in each case the number of observations is equal to 201; this is because, in the first case, the dependent variable is set to zero if the area either remained in the same quartile or worsened and, in the second case is set to zero if the area remained in the same quartile or improved. The dependent variable takes a value of one otherwise.