A changepoint analysis of UK house price spillovers

ABSTRACT

We study spillovers between regional housing markets in the UK in the period 1973–2020. The analysis is based on a vector autoregressive model that allows for structural breaks in its parameters at unknown times. In particular, we allow for distinct breakpoints in the conditional mean, variance and correlation parameters, which enables us to distinguish different spillover channels. Based on the resulting piecewise constant model, we compute the spillover index by Diebold and Yilmaz. We find significant time variation of the spillover index that indicates a decreasing role of London for the rest of the country, but that also indicates reduced contagion risk and the existence of the North–South divide that declined later in the sample. Furthermore, a central role of the Midlands is demonstrated.

KEYWORDS:

• vector autoregression
• structural breaks
• contagion
• spillovers
• regional housing markets

JEL:

• C32
• G01
• R10
• R31

ACKNOWLEDGEMENTS

The authors thank the seminar participants at WU Vienna, the associate editor and four anonymous referees for valuable comments that led to substantial improvements being made to the paper. The views expressed here are those of the authors alone and do not necessarily reflect the policies of Statistics Netherlands.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. Authors’ calculation using OECD Analytical House Price Data Base. EA-15 refers to the Euro area of 15 countries.

2. Nationwide house prices are mix adjusted, that is, tracking a representative house price over time rather than the simple average price. For details, see https://www.nationwide.co.uk/-/media/MainSite/documents/about/house-price-index/nationwide-hpi-methodology.pdf/.

3. Instead of using the population-weighted averages, we could also use principal component analysis for aggregation of the English regions as done, for example, by Yang et al. (2018). The resulting aggregated indices were basically equivalent, with a correlation of 0.99, with the weighted ones.

4.  Details are available from the authors upon request.

5. We refer the interested reader to Qu and Perron (2007) for technical details.

6. We rely on the generalized forecast error variance decomposition, as in Diebold and Yilmaz (2012), that does not identify structural shocks with a clear economic interpretation. An alternative would be the approach suggested by Yang et al. (2021). However, due to our rather small sample size, this approach is not feasible here.

7. The results for other forecast horizons, such as five or 20 steps, yield virtually identical results. Detailed results are available from the authors upon request.

8. The Association of Residential Letting Agents (ARLA) and four lenders had launched the ‘buy-to-let initiative’ in September 1996 which promoted investment in property.

9. We thank two anonymous referees for this suggestion.