Property market modelling and forecasting: simple vs complex models
| Pages | 337-361 |
| Published date | 06 July 2015 |
| DOI | https://doi.org/10.1108/JPIF-08-2014-0053 |
| Date | 06 July 2015 |
| Author | Arvydas Jadevicius,Simon Huston |
| Subject Matter | Property management & built environment,Real estate & property,Property valuation & finance |
Property market modelling
and forecasting: simple
vs complex models
Arvydas Jadevicius and Simon Huston
Department of Real Estate, School of Real Estate and Land Management,
The Royal Agricultural University, Cirencester, UK
Abstract
Purpose –The commercial property market is complex, but the literature suggests that simple models
can forecast it. To confirm the claim, the purpose of this paper is to assess a set of models to forecast
UK commercial property market.
Design/methodology/approach –The employs five modelling techniques, including
Autoregressive Integrated Moving Average (ARIMA), ARIMA with a vector of an explanatory
variable(s) (ARIMAX), Simple Regression (SR), Multiple Regression, and Vector Autoregression (VAR)
to model IPD UK All Property Rents Index. The Bank Rate, Construction Orders, Employment,
Expenditure, FTSE AS Index, Gross Domestic Product (GDP), and Inflation are all explanatory
variables selected for the research.
Findings –The modelling results confirm that increased model complexity does not necessarily yield
greater forecasting accuracy. The analysis shows that although the more complex VAR specification is
amongst the best fitting models, its accuracy in producing out-of-sample forecasts is poorer than of
some less complex specifications. The average Theil’sU-value for VAR model is around 0.65, which is
higher than that of less complex SR with Expenditure (0.176) or ARIMAX (3,0,3) with GDP (0.31) as an
explanatory variable models.
Practical implications –The paper calls analysts to make forecasts more user-friendly, which are
easy to use or understand, and for researchers to pay greater attention to the development and
improvement of simpler forecasting techniques or simplification of more complex structures.
Originality/value –The paper addresses the issue of complexity in modelling commercial property
market. It advocates for simplicity in modelling and forecasting.
Keywords United Kingdom, Market, Property, Modelling, Complex, Simple
Paper type Research paper
Property market modelling and forecasting
Property market modelling and forecasting is an indispensable activity in prope rty
investment (Mitchell and McNamara, 1997). The issue has been the subject of a number
of studies. As a result, numerous models have been developed to forecast proper ty
markets (Brooks and Tsolacos, 2010). According to Harris and Cundell (1995, p. 76),
“the market crash which traumatised the property industry between 1991 and 1994 has
led the institutions in particular to seek greater predictive input to their portfolio
management and investment decisions”. As McDonald (2002) points out, after the
1980s property boom property researchers responded to the crisis situation, and as a
result substantial progress has been made in property market research and forecasting.
Though Tonelli et al. (2004, p. 1) argues that “numerous econometric models have
been proposed for forecasting property market performance, but limited success has
been achieved in finding a reliable and consistent model to predict property market
movements”, property researchers, including McGough and Tsolacos (1995), Wheaton
et al. (1997), Barras (2009) and Bork and Moller (2012) suggest that the property market
is forecastable.
Journal of Property Investment &
Finance
Vol. 33 No. 4, 2015
pp. 337-361
©Emerald Group Publis hing Limited
1463-578X
DOI 10.1108/JPIF-08-2014-0053
Received 5 August 2014
Revised 22 December 2014
2February2015
Accepted 13 March 2015
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1463-578X.htm
337
Property
market
modelling and
forecasting
Simple and complex models
According to Caminiti (2004, p. 992), models are “invaluable tools”. Models help users
to develop a better understanding of complex systems, allow testing for possible
scenarios, predicting outcomes, as well as they can assist in the setting of priorities.
Byrne et al. (2010) add that models have been produced for a range of different reasons
(i.e. to improve one’s understanding on the subject and its processes, to predic t, forecast
or explore possible scenarios, or to provide a basis for decision making).
Despite the benefits of models, many concerns are expressed regarding their
application. STOWA/RIZA (1999) observes that extensive use of models increases the
risk of inexpert use which as a result can lead to unreliable modelling outcomes.
Similarly, Middlemis et al. (2000) comments that if the model is poorly designed,
or it does not represent the system being modelled properly, all efforts to create the
model are virtually in vain, or are likely to generate inaccurate forecasts. Subsequently,
Jakeman et al. (2006) notes difficulties associated with models. According to the
researcher, the use of models can bring unwanted outcomes due to “limitations,
uncertainties, omissions and subjective choices in models”( Jakeman et al. ,2006, p. 603).
The issue of model use and application is also addressed by Box and Draper (1987),
Sterman (2002), and Mellor et al. (2003), to name but a few. These commentators
suggest a more rigorous critique of the subject. According to Box and Draper (1987,
p. 424), “essentially all models are wrong, but some are useful”. Following Mellor et al.
(2003, p. 16), “models offer more hindrance than help”. As Sterman (2002, p. 525) states,
“all decisions are based on models, and all models are wrong”. For Sterman, failure is
built in models as they are only a simplification, an abstraction of the system at no solid
foundation. Moreover, Sterman claims that models are based on unreliable human
perception and knowledge. However, despite all of this criticism, researchers including
Parker et al. (2002) and Caminiti (2004) consider models to be assets rather than
liabilities and essential elements in understanding and forecasting complex systems.
What is the difference between simple and complex models?
Chorley (1967) was perhaps the first to generate a modelling typology , but within
sequential forecasting arena few direct comparisons of simple and complex models are
available. What is more, the distinction between a simple or complex remains obscure
(Buede, 2009). Many researchers, including Armstrong et al. (1984), Armstrong (1986),
Wilkinson (1999), and Sterman (2002) refer to simple and complex models without
providing working structural definitions.
Batty and Torrens (2001), on the other hand, define a complex system as, “an entity
which is coherent in some recognisable way but whose elements, interactions, and
dynamics generate structures admitting surprise and novelty which cannot be defined
a priori”(Batty and Torrens, 2001, p. 2). According to Holland (1995), a “complex”or
“adaptive”model is one that maintains its composition and coherence through time.
In case of the socio-economic environment, Allen and Strathern (2005) suggest that a
simple model is a structure of fixed, predictable behaviour, while a complex one is a
system in which a range of possible structural changes can appear. According to Buede
(2009), complex, or as the author indicate, science-based models, usually require greater
amounts of data, as well as a greater set of relationships.
In property, as in other fields of research, only minor references as to what constitute
simple and complex models are discussed. According to Chaplin (1999), econometric
models, which are considered as complex structures, differ from simple competitors as
they include more variables and contain a greater number of estimations. Brooks and
338
JPIF
33,4
To continue reading
Request your trial