Hear our experts’ take on the latest developments and trending topics
The importance of forecasts
Every well-planned real estate investment decision is underpinned by a series of market and site-specific forecasts. For a construction project, the most important are almost inevitably the projected sales price and schedule, and construction costs, whilst for an investment, it is future rents, eventual sales price, yields, and most probably, financial variables such as interest rates. We hear a lot about feasibility studies, but what a good feasibility study really amounts to is a series of accurate forecasts supported by an explanatory text.
The results of poor real estate forecasting are all around us. Failed developments and unwise investments, business disruption, and the ultimate bankruptcy of individual developers are all the price to pay for forecasting failure. Real estate consultancies need to understand how their experts are producing the numbers on which development and investment decisions are being made.
Forecasting models of the UAE market, as for any other real estate market, broadly divide into two types. On the one hand there are time series models, which rely on historical data, such as the Box-Jenkins Autoregressive Integrated Moving Average (ARIMA). Time series models work best in stable markets where there are no major expected changes in market conditions.
The second type, regression models, rely on a series of independent variables which are analysed as a combination explanation of a dependent variable, such as UAE house prices. The independent variables are then separately projected and then combined to produce a forecast of the dependent variable. Hence, ARIMA can be modified and combined with a vector of an explanatory variable(s) (ARIMAX), and this, along with Simple Regression (SR), Multiple Regression, and Vector Autoregression (VAR) are all examples of this approach. VAR models, for example, include more than one dependent variable and the independent variables include lags of the dependent variable from each of the equations in the model. Typical independent macro-economic variables are demographics, interest rates, GDP growth and levels of FDI, and in the case of the Abu Dhabi market to a larger degree than in Dubai, the oil price as well. Market variables include the supply pipeline and sale clearance data. All the evidence suggests that bigger, more complex models do not necessarily have greater predictive power. What really matters is ensuring that the independent variables chosen are really those driving the market, and that they will continue to do so.
The Cinderella of forecasting techniques, which is often used in the UAE, is to rely on the opinions of experts. There is nothing necessarily wrong about using this kind of qualitative technique. However, it is as well to recognise that there is a world of difference between taking the advice of one expert, who may be parti pris, if they are, say, hoping for more work if the project goes ahead, and a coherent qualitative forecasting system such as Delphi, which although can and has worked very well in the region, require a formal ‘focus-group’ setting and a modicum of actual organisational skill to be effective. Most forecasters already do use qualitative overlay processes to move away from econometric model estimates, citing mood and sentiment, but at the cost of introducing subjectivity into their forecasts, as a result. This tendency has accelerated since the financial crisis, as models produced before then had a ‘business-as-usual’ component to them which was not reflected in parallel discussions of the likely trajectory of the market.
Evidence suggests that it is easier to forecast rents rather than yields, and that construction costs are still less of a problem. The main reason for this is that construction is essentially a short-term issue, and whilst there are exceptions, costs can be constrained through contract arrangements and stockpiling. The most difficult forecasting task of all is how to predict land prices in the medium and long term, which is an especial concern given the overall trend of rising land prices by comparison to construction costs throughout the UAE and worldwide.
Forecasters have also now stressed the importance of a global perspective on real estate markets, which certainly applies to the UAE. An appreciation of relative currency performance given the US dollar peg of the dirham, global commodity prices, and trade relationships both within the region and worldwide, are all exogenous factors which should make an appearance in properly constructed real estate regression models.
The final point to note is that there is strong evidence to suggest that ‘by combining forecasts from different methods and sources, greater predictive results can be achieved’. Certainly, wise users of UAE market forecasts never rely on the output of one model, one methodology, or even one provider.
What is said publicly about methodologies is still relatively limited. For example, in their outlook on the real estate market in 2020, PwC said they had predominantly focused on GDP, based on the assumption that real estate will continue to form a consistent percentage of real GDP in the future. The difficulty of applying this to a market such as the UAE where economic diversification will gradually bring that percentage down is obvious.
Sometimes it is very easy to understand why firms are reluctant to publish their methodologies. Early in 2018, ANZ Bank revealed that their new model suggested the Australian housing market would continue to rise, albeit modestly, on the back of an Error Correction Model (ECM). The theoretical benefit of ECMs is that they explicitly highlight the market’s role to remove demand and supply imbalances resulting in a market equilibrium. The ANZ model included an equation that included a ‘revert to trend’ term, the lagged growth rate of house prices, changes in housing investment, gross incomes, and the average mortgage interest rate. The model had excellent retrospective correlation properties – i.e. it explained the past very well indeed. It however failed to predict the dramatic fall in Australian property prices over the past year.
When a regression model of this kind fails, the answer is certainly not to throw up one’s hands and say, if an organisation with the resources of ANZ Bank and with every reason to seek accuracy in its view of the market can get it so wrong, what chance do we have? It is to ask carefully where the model went wrong, and learn from the experience.
Sometimes it is a failing of data: in the past, data for the UAE was inadequate, but with services such as Property Monitor now available, there are now plenty of real estate data to use. Macroeconomic statistics are equally important, and these are also now available in long time series. In this case, though, it looks strongly as if auto-regressive trend modelling ended up producing a ‘business-as-usual’ result which did not incorporate the effect of continued apartment oversupply, political threats to generous tax breaks, and a flat macro-economic outlook for which market participants were unprepared. So, whilst there are solid empirical methods to evaluate the accuracy and usefulness of forecasts, including mean squares, geometric mean, relative absolute error or mean absolute error, these can only tell a user how effective a forecast has been in the past. If what influences real estate markets changes, and those changes are not captured in a model, then the result will inevitably fail, in turn, to make an accurate prediction of the market.
Lessons for the market
Despite its obvious importance, forecasting is still regrettably regarded as an arcane subject. Too many developers and investors still ‘pay for the pdf’ and fail to pay sufficient attention to the methodologies their consultants have used, and the assumptions on which the output presented to them has been based. These are the key questions that should always be posed when the results of a model are presented.
In part, this can be explained by the practical difficulties many real estate companies have in postponing decisions or cancelling planned projects when the numbers don’t stack up, but in the post-financial crisis world where we are all very much aware of the consequences of poor forecasting, the importance of getting over those difficulties and making decisions based on the best possible view of the future is now increasingly accepted. The time when the kind of issues raised here can be ignored by investors, banks, developers and investors is over – and that is a good thing.
 Hepsen, A. and Vatansever, M. (2010) Forecasting Future Trends in Dubai Housing Market by Using Box-Jenkins Autoregressive Integrated Moving Average International, Journal of Housing Markets and Analysis, 4(3), 210-223
 Jadevicious, A. and Huston, S. (2015) Property market modelling and forecasting: simple vs complex models. Journal of Property Investment & Finance, 33(4). 337-361
 Pivo, G. (2008) Responsible property investment criteria developed using the Delphi Method. Building Research and Information 36(1), 20-36.
 Jadevicius, A., Sloan, B. and Brown, A. (2012) Examination of Property Forecasting Models –
Accuracy and its improvement through Combination Forecasting, p.1. Available at: https://eres.architexturez.net/system/files/pdf/eres2012_082.content.pdf Retrieved 25 July 2019.
 PwC (2014) Real Estate 2020
Building the future. Available at: https://www.pwc.com/sg/en/real-estate/assets/pwc-real-estate-2020-building-the-future.pdf Retrieved 25 July 2019.
 Available here: Yardley, M. (2018) There’s now a new model for forecasting Australian housing prices – and it works! propertyUpdate.com.au (blog) 4 March 2018, Available at: https://propertyupdate.com.au/new-model-for-forecasting-australian-housing-prices/ Retrieved 25 July 2019.
Stay up to date
|_ga||2 years||The _ga cookie, installed by Google Analytics, calculates visitor, session and campaign data and also keeps track of site usage for the site's analytics report. The cookie stores information anonymously and assigns a randomly generated number to recognize unique visitors.|
|_ga_34E12VSHW6||2 years||This cookie is installed by Google Analytics.|
|_gat_gtag_UA_66458947_1||1 minute||Set by Google to distinguish users.|
|_gat_UA-66458947-1||session||A variation of the _gat cookie set by Google Analytics and Google Tag Manager to allow website owners to track visitor behaviour and measure site performance. The pattern element in the name contains the unique identity number of the account or website it relates to.|
|_gid||1 day||Installed by Google Analytics, _gid cookie stores information on how visitors use a website, while also creating an analytics report of the website's performance. Some of the data that are collected include the number of visitors, their source, and the pages they visit anonymously.|
|cookielawinfo-checkbox-statistics||1 year||This cookie is set by the GDPR Cookie Consent plugin to store the user consent for the cookies in the category "Statistics".|