A non-parametric CAE approach to office rents: Identification of Helsinki metropolitan area submarkets

https://doi.org/10.1016/j.eswa.2011.07.037Get rights and content

Abstract

The study attempts to identify and estimate the office rents of submarkets in the Helsinki metropolitan area. We applied a non-parametric empirical approach called the CAE method to identify six parameters: highway APD (access point distance), car traffic density, light rail APD, main retail distance, office building density and effective age. Our results suggest that car traffic density is the single most influential parameter. Office rent decreases with effective age and increases with the density of office buildings. Longer distances to highway access points and to the main retail centres decrease office rents, while shorter distances to the light rail access points increase office rents in general and particularly for locations close to highway access points. We identified local peaks by inspecting multiple graphs. The local peaks were considered evidence for the existence of commercial office submarkets within the Helsinki metropolitan area. We identified seven submarkets at different rent levels. Interpreting submarkets from the CAE graphs allowed us to recognise particular business districts in the Helsinki metropolitan area. In addition, it is of great significance that the roles of the given and estimated variables can be exchanged. The method is directly applicable in real estate studies using adapted database and prescribed smoothing parameters.

Introduction

In the domain of commercial office space, firms consider the advantages and disadvantages of office building quality, neighbourhood development and location according to their customers’ and employees’ needs and perceptions. Among these, location comprises a complex phenomenon depending on various factors detectable by market players. Researchers have tried to detect the location factors, to predict their value using various approaches and to determine their different levels of influence using scientific approaches.

In examining the relationship between a city’s spatial impact and its market for office space, pre-existing geography is an inertial force in creating a path dependency for the location of office-based services. The establishment of a large market for office space can create agglomeration economies (Lizieri, 2009). Early (Clapp, Kim, & Gelfand, 2002; Clapp, Pollakowski, & Lynford, 1992), and recent (Brounen and Jennen, 2008) studies suggest that spatial patterns are strongly related to market dynamics. Brounen and Jennen (2009) created a measure of growth potential that allows for the evaluation of office agglomerations over time. Nappi-Choulet and Maury (2009) proposed a Bayesian STAR model to control for heteroskedasticity in the market for office space in Paris. Hysom and Crawford (1997) concluded that location is an amalgam of several factors which, when combined, impart a specific value to a location. Intra-metropolitan location was found to depend on agglomeration economics, which consist of localised economies that accrue to firms of the same sector. Agglomeration economies are represented by the density of office buildings. Profit maximisation could be enhanced by agglomeration, especially when office or commercial retail space is highly dense and in locations where there is potential for increasing the ratio of future rents to land, labour and transport costs (Jennen and Brounen, 2009, Margulis, 2007).

Several researchers have suggested that commercial firms value access to service centres (Bruinsma, 1997, Koppels, 2006, Ryan, 2005, Sivitanidou, 1995, Sivitanidou, 1996). A similar explanation and identification was reported earlier by Bruinsma (1997), who stated that the centrality and the accessibility of an office best explain why firms tend to value cities as location sites. Trevillion, Wang, and Withall (2006) reported on the shift of offices away from traditional central business districts (CBDs) and into new business locations, citing difficulties in fulfilling long-term demand requirements in the market for office space in Edinburgh. Koppels (2006) proposed an analysis of rent variations within metropolitan boundaries at a single moment in time, an approach by which the added value of office property attributes and site-specific characteristics can be assessed. Research by Jakobson and Onsager (2005) considered head offices as flow nodes and emphasised their geographical concentrations. Ryan (2005) and Ghebreegziabiher and Jasper (2008) examined the importance of access to light rail transit and highway systems in estimating office and industrial property rents. Hedonic price analysis was applied to properties in the San Diego metropolitan area and Amsterdam’s South Axis area. The results indicated that access to highways and access to light rail are significant factors in estimating office property rents.

A different approach was proposed by Mourouzi-Sivitanidou (1999). She explored adjustments in metropolitan office rental rates towards their implicit equilibrium levels. Sing, Ooi, Wong, and Lum (2006) presented results on the office space preferences of occupants of Suntec City. Based on mean scores, they revealed that prestige of an office location and accessibility by public transport are the two factors most highly ranked by firms.

Recent studies have analysed and identified office submarkets. Dunse, Leishman, and Watkins (2001, 2002) argued that there are sound theoretical and technical arguments for segmenting office markets into distinct submarkets. They suggested that office submarkets should be derived empirically, rather than according to the prior knowledge of agents or researchers, by applying principal components and cluster analysis. Among the attributes they tested were age of the real estate, distances to key points of accessibility and location quality indicators. More recently, Archer and Smith (2003) explained office clustering (i.e., office submarkets) using the nature of office functions, related industry economies of scale, and office location in the context of modern urban theory and communication requirements. McMillen and Smith (2003) claimed that large metropolitan areas with high congestion levels are virtually certain to have at least one sub-centre and that the number of sub-centres rises with population and commuting costs. These two variables account for nearly 80% of the variation in the number of sub-centres across urban areas. Many studies of property markets use Geographic Information Systems (GIS), which can enable one to better locate business clusters (i.e., Jennen and Brounen, 2009, Rodriguez et al., 1995, Thrall, 1998, Weber, 2001).

Karakozova (2004) is, to our knowledge, the only study published of the market for Helsinki office space. She investigated the variation in office capital growth over a 30-year period, from 1971 to 2001, using three alternative models: a regression model, an error correction model (ECM), and an integrated autoregressive-moving average model with exogenous explanatory variables (ARIMAX).

This paper relies on a new approach to identifying submarkets and estimating commercial office properties, dealing simultaneously with six influential parameters without any a priori assumptions. We describe and apply a non-parametric empirical approach called the CAE method. We also discuss the input variables that are most influential in determining office rents. Our important results are shown in graphs, enabling analysis of the influence of the input variables on the office rents and the identification of submarkets. This study is part of a research project conducted at the Institute of Real Estate Studies, Helsinki University of Technology TKK.

Section snippets

The CAE method

The CAE method is based on a special type of multi-dimensional, non-parametric regression and represents a kind of probabilistic neural network. Developed by Grabec and Sachse, 1991, Grabec and Sachse, 1997, it enables relatively simple empirical modelling of different phenomena. The method has already been used in the field of structural engineering to model attenuation relationships (Fajfar & Peruš, 1997) and, more recently, to model the force–displacement envelope of dynamically loaded RC

Database

The database consists of 660 records from various sources. As described in Section 1, different variables have been identified by various researchers as the most influential for determining office rent. However, in order to keep the study manageable (and to facilitate practical and graphical presentation of the results), we sought to keep the number of input parameters as low as possible. (Note that this is not a limitation of the CAE method.) After trial and error and intensive discussions, we

General trends

In this chapter, we present and discuss influences on estimated office rents as quantified by the input parameters. Office rent was estimated as a function of two different input parameters (Fig. 3) as well as all six input parameters at the same time (Fig. 4, Fig. 5, Fig. 6, Fig. 7). In the latter case, more than one graph is needed to observe the behaviour of functions in 6+1-dimensional sample space. The two input parameters are presented as continuous values on the abscissa and ordinate,

Conclusions

By interpreting CAE graphs, it is possible to recognise particular business districts in the Helsinki metropolitan area as office submarket areas, which coincide with the parametric descriptions of the submarkets. An office submarket is defined as a group of offices where, although the characteristics of each office are different, offices serve as substitutes for one another. Using multiple graphs, we identified local peaks, which were considered evidence for the existence of commercial office

Acknowledgements

The authors wish to acknowledge the National Technology Agency of Finland (Project 40079/06) for financial support.

References (35)

  • J. Clapp et al.

    Predicting spatial patterns of house prices using LPR and Bayesian smoothing

    Real Estate Economics

    (2002)
  • J. Clapp et al.

    Intrametropolitan location and office market dynamics

    Journal of the American Real Estate and Urban Economics Association

    (1992)
  • N. Dunse et al.

    The existence of office submarkets in cites

    Journal of Property Research

    (2002)
  • N. Dunse et al.

    Classifying office submarkets

    Journal of Property Investment and Finance

    (2001)
  • P. Fajfar et al.

    A non-parametric approach to attenuation relations

    Journal of Earthquake Engineering

    (1997)
  • D. Ghebreegziabiher et al.

    The effect of railway stations on office space rent levels: The implication of HSL South in station Amsterdam South Axis in Railway

  • I. Grabec et al.

    Automatic modelling of physical phenomena: Application to ultrasonic data

    Journal of Applied Physics

    (1991)
  • View full text