Synthesis: bottom-up versus top-down approaches to factors in commercial real estate

It seems that commercial real estate pricing models based on a strict interpretation of the Asset Pricing Model (macro-variables) and Hedonic Pricing Theory (microfactors) might have barked at the wrong tree. Industry-led research as well as pricing models inspired after valuation methodologies which adopt a bottom-up approach to real estate risk have managed to cover the full spectrum of real estate risk, i.e. from macro-level risks stemming from the national economy to the most granular micro-scale at the property or even unit level. Such extreme granularity has been more or less overlooked in classic models. Modelling real estate’s micro-scale is in fact very straightforward. Identifying relevant variables to be included in models is also straightforward, at least on paper. Dealing with cash flows supposes to select factors potentially affecting a building’s income-generating ability, i.e. factors stemming from what Ratcliff identified as the “legal dimension” of real estate.

Noticeably, models which include a micro-scale level of analysis tend to be very parsimonious in macro-variables. For instance, Miles et al. (1990) highlight very few macro-variables from their “national location” category in their models of the four commercial property types, notwithstanding the localness of these variables which are selected at the county level. Likewise, Crosby et al. (2016) only include three macrovariables in their optimal model (risk free rate, anticipated inflation and return on alternative investments) out of 18 variables overall. Does that mean that micro-scale variables have the ability to capture macro-trends at the asset level, i.e. where it matters for investors? A similar point can be raised about spatial hedonic models which position location as a dynamic factor within a complex eco-system whose impact is felt at the most granular level, i.e. the property and its immediate vicinity.

Due to the ever present issue of multicollinearity in factor models, being able to reduce the number of variables in real estate pricing models by fine-tuning variable identification and selection is a great advantage of a bottom-up approach.

It might suppose for real estate finance to find new roots in the valuation paradigm. The fact is that real estate finance has been aiming to develop pricing models without looking in detail at cash flows at the property level even though this information exists and could be readily accessed (under the right conditions of course). In its top-down approach searching for ‘normalities’, urban economics might have neglected an essential dimension of real estate. From investors’ viewpoint, a building is a physical structure built on a unique piece of land in order to generate income. One cannot develop a holistic view of real estate risk without a clear reference to its micro-scale components.

Richard Feynman once jokingly asked: “are bricks essential objects?” (Feynman, 1985). Surely, buildings are not abstract entities even in the perfect world of finance. In line with Ratcliff (1961), they are part of a very material process that turns land into space and ultimately money. What might negatively impact the smooth working of this process whereby the real estate sector transforms bricks into cash flows boils down to risk.

Another noteworthy point is that most real estate pricing models are developed without a clear reference to a risk model of commercial real estate. Valuation-inspired models which, de facto, refer to a conceptual framework anchored in income generation at the most granular level are notable exceptions. Logically, deriving factors from a cap rate perspective rather than a price perspective should contribute to almost naturally adopting a bottom-up approach to variations in returns, and thus risks. The absence of proper structural models of real estate risk is a concern because it means there is no certainty about the choice of variables in factor hedges. Based on academic papers mentioned in Appendices 1.1 and 1.3, one would be hard-pressed to select a small number of macro- and micro-variables, respectively, as underlyings to factor hedges among the myriad mentioned in past studies. Nonetheless, this is what was done in the replication and hedging study of London office buildings reported in Appendix 1.4, out of necessity rather than choice as the desired variables were not publicly available.

This is the paradox of applying a theory designed for financial assets (such as APT) to real estate assets. A top-down approach which ignores the actual nature of commercial real estate as a quintessential!}' heterogeneous asset class results in anecdotal empirical evidence very far from the normalities the theory would have needed to be fully vindicated in the odd context of real estate. Even the application of ever more sophisticated econometric techniques cannot hide the fact that very little generalisable knowledge comes out of a study which is not underpinned by a model of commercial real estate risk. The following section covers the need for risk models in commercial real estate as a prerequisite to designing relevant and successful property derivatives.

 
Source
< Prev   CONTENTS   Source   Next >