The determinants of European office rents

Real estate is a complex economic commodity – it is durable, spatially immobile, and has limited divisibility. These characteristics mean that the structure and operation of the market is more complex than that of many other assets. For instance, while the inherent business needs of occupiers are met by real estate space, the durability of the commodity means that it also exhibits many of the characteristics of an investable asset. In this context, even at the simplest level, real estate economists tend to conceptualise the market as consisting of occupier and investor markets. More advanced analysis extends and develops this to recognise the distinctive nature of the supply side of the market and/or to consider the importance of spatial differentiation or to segment the market along quality lines. This makes real estate markets analytically complex and difficult to model. The desire to find effective models that can predict or forecast outcomes reliably and inform the decisions of users, developers and investors has given rise to the development of a large number of models over the past forty years.

Econometric models of the UK and European office markets form a large part of this empirical literature, and these models tend to be dominated by demand-side considerations. This approach is an appropriate method to understand the occupant submarket at least in the short term, as the rental determination process is clearly described in the single equation model and supply responsiveness is very limited. The rigid short-term supply side, however, is not generally considered explicitly, and any explanation of dynamics, temporary disequilibrium and convergence to the new equilibrium is missing.

The modelling literature divides into two broad groups: the first group consists largely of single equation models of rent that combine demand-side variables with limited consideration of the supply side; the second group consists of multi-equation models that seek to take more explicit account of the complex structure and dynamics of the market. For instance, some of the more complex modelling projects start with the textbook treatment of the office real estate market, generally referred to as the DiPasquale-Wheaton (DW) model. In the DW framework, the dynamics of the market is described as an anti-clockwise operation process that goes through ‘submarkets’ or quadrants. It begins with a short-term rental adjustment process of demand and supply, continues to price valuation of property, stimulating construction starts, and then moves to the level of total space stock adjusted by new supply, depreciation and demolition. In this framework, rent is highlighted as an occupation indicator in the short term, and a transmission factor which brings user sector change to other submarkets. In this context, the four-quadrant model and relevant discussions on the interactions of submarkets justify the adoption of a multi-equation approach in the analytic perspective of real estate economics. Since multi-equation models include subsectors as component equations and estimates simultaneously, the methodology enables comprehensive analysis in reflection of the theoretical interdependency.

Relevant model-building studies have tended to be more common in the U.S. than in the European market. Even though there are a few European analyses, they have tended to address one specific variable, single metropolitan or country market, or a subsector (e.g., the user market) rather than comprise comprehensive, systematic market analysis. Against this backdrop, this study aims to focus on the interaction process of submarkets and their convergence.

Based on the above, this study seeks to explore the determinants and dynamics of commercial real estate, particularly office rents, and apply these to the European office market from a neoclassical economic perspective. The interaction process of subsectors within the market system and their spatial integration impacts that go beyond individual cities or countries will be further examined. To achieve this aim, three objectives are established: (1) to understand the determinants of office rents in selected European markets, (2) to explore differences across space and over time in the selected markets (space dynamics), and (3) to examine the interactions among the subsectors of the office markets.

In this study, the sample consists of six office centres, namely, London, Paris, Frankfurt, Amsterdam, Madrid and Milan, for the period 2007 to 2018 on the quarterly basis. The modelling approach starts with a simple operational framework and then seeks to add complexity, reflecting the DW theoretical model and the interaction it depicts. In the theoretically richer variants, each of the user, investment, flow and stock supply sectors (quadrants) are considered. Simultaneous estimation of the four equations enables the application of theoretical interaction of subsectors in the regression process.

The modelling work assesses different model structures and employs a range of methods of estimation. This allows for a systematic evaluation of the performance of the models in terms of their theoretical consistency, model properties and explanatory power. The role of endogenous variables needs to be highlighted, since a variable is not only relevant as a component of one equation but delivers change to the adjacent subsector equation as well. For instance, endogenous variables in the investment and development sectors indirectly impact on rent through the error term, while predictors in the occupier sector (rental equation) directly impact on the dependent variable (rent). The process demonstrates the structural interaction between equations, and implies that simultaneous estimation performs better than a single equation since additional information is reflected in the system.

There are significant technical challenges associated with the empirical analysis. Despite the theoretical advantage of more complex model structures, simultaneity is a significant econometric issue in the estimation. Simultaneity accompanies endogeneity of certain predictors, and the resultant heterogeneity causes the Gauss-Markov assumption to be unsatisfied. Since ordinary least squares (OLS) is no longer the best linear estimation unbiased estimator (BLUE), alternative methods, such as robust general least squares (GLS), panel, and three-stage least squares (3SLS), and generalised method of moments (GMM), are estimated. Each method improves OLS, as follows: 1) Robust GLS justifies the BLUE violation; 2) panel analysis enables regression of cross-section combined with time-series; 3) 3SLS reflects intersectoral endogeneity with the simultaneous estimation structure; and 4) GMM compares the maximum likelihood estimation (MLE) performance to that of OLS.

This research study sheds light on the workings of European office studies by 1) narrowing the gap between demand-supply equilibrium theory and the empirical approach, and 2) improving the quality of econometric estimation by expanding the scope of analysis. First, the four-quadrant mechanism worked well in 3SLS estimation through its component equations. The equations are simultaneously estimated, and endogenous variables delivered the impact of one subsector to the others. The actual coefficients are generally matched with expected signs. Although some of the supply-side variables are statistically insignificant, the interaction processes properly operate in the comprehensive demand-supply system. Second, panel and 3SLS present improved estimation methods in terms of the analytic dimension. In the regression process, multiple office centres and submarkets are combined with time-series in the panel analysis and 3SLS, respectively. Comparison of both results to OLS supports the hypothesis that the quality of estimation is enhanced by structurally combined information.