Eigen analysis of the stability and degree of information content in correlation matrices constructed from property time series data
Volterra Consulting Ltd 5, The Old Power Station, 121 Mortlake High St, SW 14 8SN
Corresponding author: a firstname.lastname@example.org
Published online: 15 May 2002
Property is an asset which forms part of the portfolios of many investors, particularly institutional ones, along with equities and bonds. Techniques from physics, particularly that of random matrix theory, have provided powerful insights into the behaviour of financial assets. A large database providing time series data for over 10,000 individual properties is available for the UK. Some of the data is available at an annual and some at a monthly frequency. However, even at the monthly frequency, only a relatively small number of observations is available, certainly in comparison with that available with financial assets. A key issue in translating methods of analysis in financial markets to property data is whether they are applicable given the small number of data points available. This paper addresses this issue. Using the tools of random matrix theory, we find that a great deal of information is contained within property data. The correlations between different types and geographical locations of property tend to have far more true information and be more stable over time than is the case with financial data, despite the large number of observations available with the latter.
PACS: 01.75.+m – Science and society
© EDP Sciences, Società Italiana di Fisica, Springer-Verlag, 2002