The engagement of the broader investment community is ever more essential to forecasting in today's uncertain market conditions. Henri Vuong and Ruth Hollies report

In an uncertain environment, the success of any strategy is highly dependent on the ability to foresee and prepare for the future. Forecasting is a vital tool in this process. This article explores the value and difficulties of the forecasting process, and identifies where commercial property investors can make improvements.

The process of modelling generally begins with postulating which variables may be needed before data gathering, followed by exploratory data analysis, identifying relationships (contemporaneous or lagged), specifying an appropriate model, fitting the model and then robustly examining the model outputs.

Models provide structural frameworks for understanding market dynamics, helping to quantify historical relationships between market indicators (such as supply and demand drivers) and performance measures. Studies have shown that past behaviour can provide useful insight into future market behaviour.

A sound understanding of such relationships, combined with scientific analytical approaches, can help project prospective performance and greatly assist the investment and decision-making processes. As such, forecasting has become an integral part of real estate investment, with theory-based models offering a structured framework to which the ‘ebb and flow of the market tide' can, to a substantial extent, be statistically determined.

With current market uncertainty, participants are, more than ever, casting around for views on future asset performance. Forecasting is being called upon to fill this ever growing appetite for market views but is being tested harder than ever by the violent swings in markets; forecast accuracy is even more crucial in today's uncertain environment.

Improved data availability
The availability of property market data has improved immensely, expanding in coverage and quality. For example, Investment Property Databank (IPD), which originally provided investors with annual performance data for the UK market, has expanded to quarterly and monthly UK data, and extended coverage to major global real estate markets. Similarly, as international brokerage firms have grown, so has worldwide data coverage, benefiting from local agents gathering data.

Not all plain sailing...
Forecasting models, although often central to the strategic planning process, face a variety of problems. The old adage, ‘rubbish in, rubbish out' is particularly apt when directed to the forecasting process. The technical aspects of modelling in particular are heavily reliant on sound data entering the process and consequently forecasting is reliant on the quality and quantity of forecast inputs. Data uncertainty is an inherent characteristic of any forecasting model, and can lead to spurious relationships and erroneous forecast outputs. Such uncertainty derives from various sources, most commonly measurement error, unrepresentative samples and outdated sources.

The quality of any forecast also depends on many other, more technical factors, including, but not limited to, model specification, length and nature of the time series, sample size, and whether measured relationships are lagged or contemporaneous.

The issue of uncertainty over the quality of data has a further dimension in the property market, namely, it is exacerbated by issues of data ‘inconsistency'. Variations in measurement practices between (and even within) countries lead to inconsistent measures of market indicators. For example, rents can be quoted for gross letting area or net, or headline or net effective; the meaning of net varies considerably. Similarly yield definitions differ remarkably too.

Despite much improved quality of market data and important industry initiatives towards harmonising data standards and definitions, there remains sufficient variation in the collection, measurement and construction process of key property variables for users to treat property market data with caution. Additionally, property data series are often characterised by few data points with long cycles, limited series of variables, low frequency, lagged data and valuation smoothness - making it even more difficult to find indicators that capture all these traits.

Forecasting and modelling
Modelling is distinct from forecasting. The modelling process aims to capture historical major underlying relationships, focusing on model building (selecting appropriate dependent and explanatory variables) and model checking (ensuring the model and its outputs are statistically valid).

However statistically ‘robust' a model appears to be there is always the possibility that results are spurious, which would be highly dangerous if used to forecast. As such, ‘pure model' forecasts need to be intelligently assessed and ‘sense' checked - usually reviewed by expert users. As such, the forecasting process encompasses the modelling process but augments it through the engagement of the investment community to qualitatively bring in knowledge that the models might fail or be unable to capture.

Before making such adjustments modellers and forecasters must have a comprehensive understanding of the market and its indicators to be able to confidently interpret modelled outputs. Although qualitative adjustments can improve the predictive power of models, subjective views also need to be rigorously examined to assess whether there has been any improvement.

The theory underpinning the modelling process will undeniably evolve and improve with time. However, forecasting will never be perfect, but understanding the model inputs and the measurements is a step closer to understanding the problem. As the modelling process begins with a set of prior views on how a market operates it is essential for modellers to understand the market dynamics correctly: the lag structure, the construction process: time taken from planning application to start and to completion and the impact on the market. Market level data might not capture these local-level idiosyncrasies.

Forecasters being away from the coalface of property markets might be somewhat detached from reality on the ground. In a fast-moving market like today's when valuations tend to lag transaction values, forecasters need to engage with the ‘guys on the ground' to capture these discrepancies.

A call to action
There has been rapid development in the field of real estate forecasting and the tools of the property market forecaster. Forecasting is just one element of the real estate investment process. Improvement in the quality of data can help with this process in the future, but cannot address data imperfections of the past.

To improve investor insight into future prospects for international markets, the real estate industry must now work harder to identify, understand and, where possible, address the weaknesses in current property market data and the imperfections these introduce into model forecasts - not least because measurement errors made in the past will inevitably create forecast errors for the future.

To make any meaningful inferences from forecasting models, the foibles of the data being used and the limitations of the model itself must be thoroughly understood and accounted for. Forecast outputs can be enriched with knowledge from market participants, and often the more accurate forecasts combine dispassionate quantitative model-based analysis with qualitative adjustments, reflecting market knowledge and elements to correct data issues. It is not uncommon for forecasters, irrespective of industry, to adjust model outputs in order to incorporate qualitative or other factors, more commonly known as ‘add' factors.

However, market participants should not overlook the importance of their own understanding of property market fundamentals and dynamics, hence they should not rely solely on the technical framework of the modelling process. As such, the investment community needs to be engaged in the modelling and forecasting process, bringing together the top-down strategic forecaster view with the equally important bottom-up view of market participants. Marrying and finding a balance between the technical framework and investor insight is clearly the optimal
position for many.