“Imagine a world … where we can be more certain of what the future holds for us … this report explains what you need to do to start converting fantasy into reality.”
This ambitious claim prefaces a document from the RICS Research Foundation, setting out an approach to modelling office markets. This complements the main report of a project sponsored by the Corporation of London and the RICS Research Foundation, to develop a model of the City of London office market which is presented as a portable framework for adaptation to other urban office markets.
I suppose I should declare an interest of sorts in this project and its results, in that I was once invited to join a consortium of researchers that originally tendered, unsuccessfully, to undertake this project. Naturally, my comments are now entirely without prejudice. The successful team comprised a heavyweight line‐up of forecasting and modelling expertise, including Neil Blake and colleagues from Business Strategies Limited, George Matysiak (CB Hillier Parker and City University) and Colin Lizieri (University of Reading).
The aims of the study were, first, to see if it was possible to deliver a robust forecasting model of the UK’s most data‐rich office market, namely the City of London, and second, to examine the degree to which the approach could be replicated in other centres. The bulk of the main report is taken up with development of a forecasting model for the City market.
The report devotes considerable attention to data issues. Even in the supposedly well‐researched City market, these still pose problems, notably with respect to accurate employment and office stock figures. The approach adopted largely dispenses with the “intermediate” market variables of office‐takeup, availability or vacancy rates. Rental levels are measured solely by the net effective “prime” rent. The key drivers of rental change are employment growth on the demand side and development starts on the supply side.
Market adjustments are modelled through a set of nine linked equations using half‐yearly data for varying periods, starting in 1978‐1982 through to 1998. The main exogenous driver of the model is office‐based employment in the City. On the supply side, the key variable is the level of development starts which is modelled as a function of unimplemented planning permissions, the level of real effective rents and expectations of capital growth. The last of these three is itself based on expected rents and expected gilt yields. In turn, expected rents are based on current employment change and gilt yield movements, while expected gilt yields are a reflection of changes in the inflation rate. Development completions are a lagged function of starts.
Within the model, changes to supply are largely a lagged function of rents, with planning policy influential through the stock of unimplemented consents. Within this framework, increased City employment meets inelastic supply and pushes up rents, which triggers a higher level of development starts. Lagged completions mean that large rental fluctuations are possible before new supply comes onstream and the market “clears”, at lower rents than in the aftermath of the original demand and boost.
This is, of course, a fair description of the boom‐bust cycle in the City market from the mid‐1980s to the early 1990s, when the impact of the late 1980s development boom was compounded by a sharp fall in City employment in the early 1990s recession.
The model presented is logically structured and technically sophisticated, although whether it offers significant new insights into the dynamics of the City market is debatable. It builds upon previous work using a multi‐equation approach to estimating supply and demand by Hendershott et al. (1999). The authors argue that this approach has advantages over that of Wheaton et al. (1997) in a number of respects, including a direct link to the capital markets through the influence of gilt yields on development.
The report makes no direct reference to other approaches to modelling and forecasting office markets. Much published work in this field relates to US markets. In the UK, much of the forecasting effort relating to London’s office markets, and others, has been undertaken in research consultancies and commercial property advisors. Understandably, as most of this work is not in the public domain, it is not reviewed in this report, but it should not be assumed that other approaches do not exist that may suit the purposes of forecast users equally well or perhaps better than the model in this report.
As to the particular merits of this model, and the underlying approach, I am not in a position to offer a technical critique. However, in my view, the results of this project do not quite measure up to the expectations raised by the opening quotation from the RICS Research Foundation. This applies both to the model’s treatment of the City market and to its wider applicability.
With regard to the City market, the approach invites reservations on several counts. These basically relate to various inherent difficulties with the modelling of urban office markets and how these are dealt with in this study. First, as the authors admit, the City market is modelled as a completely closed system, with no interaction with or influence from any other part of the London office market. No allowance is made, for example, for the growth in the Docklands office market since the 1980s. Many market participants would believe that Canary Wharf has had an impact on City rents over the past decade.
Second, while there is a clear case for using employment as the predominant demand driver in such a model, the questionable reliability of the employment data on which the model and any forecast depend is a problem. The authors themselves admit to doubts over the quality of the historic employment data: “… something happened in the late 1980s that our price/demand model cannot explain when using data based purely on official sources” (p. 23). Their solution is to use a dummy variable to deal with the discrepancy.
This data problem is not, of course, a failing of the study itself, but it faces any attempt to forecast employment in the City when the official statistics, or estimates derived from them, appear deficient. As a recent case in point, the ONS data for London employment in finance and business services currently show growth of less than 1 per cent in 2000, a year which saw booming office demand and record take‐up in the City market,
Third, planning policy is a key influence on office development but, as the authors note, “… market participants do not necessarily have a good way of forecasting future planning policy” (p. 19). To run the model to produce forecasts, however, requires predictions of the floorspace included in planning consents that will be granted by the City Corporation in future years.
Fourth, the model inevitably strongly reflects the historic experience of the last major market cycle in the City. Office development activity in the late 1990s was very different both in scale and in nature to the 1980s boom. A high proportion of development over recent years was pre‐let before it started. It is implicit in the model that whether development is speculative or pre‐let make no difference to its effect on supply, presumably on the basis that moves to pre‐let offices will release an equivalent amount of floorspace back to the market. This may not be wholly realistic. Such released space may be of lower quality and/or may become available after an interval of time; for example, after refurbishment. The higher proportion of pre‐let development now characteristic of the City market means the model’s predictive power for the level of development starts may be reduced.
Arguably, these are no more than quibbles about the approach taken in an inherently difficult exercise. For many potentially interested users, the real test of a forecasting model is whether it produces understandable and plausible forecasts on given assumptions. Perhaps the biggest disappointment of the report is that it provides no actual forecasts, only simulations of re‐written history showing how the model responds under various “what if?” scenarios.
The sponsors would say that the real purpose of the model is to provide a tool for others to use as widely as possible. The model can be downloaded from the RICS Research Foundation Web site for this purpose. Having done so, I would be very interested to hear from anyone who has managed to run the model obtained in this way.
The authors and the RICS Research Foundation set great store by the wider applicability of the model – “the true prize”. Sadly, the report simply affirms a belief that it will work elsewhere, although accepting that the City market may be unique in various respects that will not be replicated elsewhere and that there will be extreme difficulties in assembling adequate data to run the model for some other markets.
The report therefore includes a chapter on the portability of the model that is “in the absence of imperatorial evidence … somewhat speculative” (p. 29). This is, to put it mildly, an understatement and the report falls well short of its aim of examining the degree to which the approach could be replicated in other centres.
The accompanying RICS Research Foundation document is unequivocal: “We know that it can work” (p. 2). In the absence of a convincing demonstration that this is the case, judgment must be reserved on how far this project has helped to convert anyone’s fantasy into reality.
