Panel Data Modelling of Prime Office Rents: A Study of 12 Major European Markets

Abstract

A significant amount of real estate research has been directed towards developing empirical models explaining rental growth. This paper develops an error correction mechanism (ECM) model which is built on the general theoretical formulation of the Hendershott et a.l (2002a). Income, as measured by local market services output, is used as a key determinant of office demand. A total office stock variable is also used which is derived from the cumulated level of development completions. The model is estimated empirically for 12 office market locations across Europe using panel data techniques and the full sequence of panel model selection tests. The approach allows the behaviour of different markets to be readily compared and contrasted, and provides inferences about intra‐market dependence and the comparative speed of adjustment towards long‐run equilibrium. The study also compares the short‐term predictive performance of the proposed model to a number of alternatives which has been covered in the literature, including a‐theoretical models. The results offer some support for the proposed model for predicting short‐term rental movements.

Keywords:

• Commercial rents
• real estate modelling
• forecasting
• office rents
• real estate investment

Acknowledgements

An early version of this paper was presented in the 9th conference of the European Real Estate Society, Helsinki, June 2003, under the title ‘Structural variation in commercial property occupier markets – a panel data study of European office centres’. The authors are grateful to DTZ Research for the provision of the data used in this study. The opinions expressed in this paper are those of the authors alone. The authors are also grateful to the referees for their persistent but constructive criticism and valuable contribution in developing this research.

Notes

1. GLS is inconsistent, see e.g. Greene 1993 p. 479). However, the high statistical significance of the tests contradicts the low power of this test, enforcing the outcome that Fixed Effects is the appropriate model

2. Hsiao (1996, p.136) repeats for this case that the choice between the two types depends on whether we are making inferences conditional on the individual sample or unconditional on the population characteristics.