Monte Carlo simulations for real estate valuation
| Published date | 01 March 2006 |
| Pages | 102-122 |
| Date | 01 March 2006 |
| DOI | https://doi.org/10.1108/14635780610655076 |
| Author | Martin Hoesli,Elion Jani,André Bender |
| Subject Matter | Property management & built environment |
Monte Carlo simulations for real
estate valuation
Martin Hoesli, Elion Jani and Andre
´Bender
University of Geneva, Geneva, Switzerland
Abstract
Purpose – To address formally the issue of uncertainty in valuing real estate.
Design/methodology/approach – Monte Carlo simulations are used to incorporate the uncertainty
of valuation parameters. The probability distributions of the various parameters are constructed using
empirical data and a simple model is suggested to compute the discount rate.
Findings – The central values of the simulations are in most cases slightly less than the hedonic
values. The confidence intervals are found to be most sensitive to the long-term equilibrium interest
rate being used and to the expected growth rate of the terminal value.
Research limitations/implications – Further research should focus on the stability of the model
when other portfolios are used and for different periods of the real estate cycle. It would also be fruitful
to dig deeper in the relation between capital expenses and property values.
Practical implications – Risk can be assessed by valuers as they can measure the probability that
the value of a property be less than a given threshold.
Originality/value – By incorporating uncertainty, the analysis does not yield merely a point
estimate of the property’s value but rather the entire distribution of values. Also this paper constitutes
a contribution to the debate about valuation variation and the margin of error in valuing properties.
Keywords Monte Carlo simulation,Real estate, Net present value,Uncertainty management
Paper type Research paper
1. Introduction
Among the various approaches to valuing real estate, the discounted cash flow (DCF)
method, using the weighted average cost of capital (WACC) as the discount rate, is well
accepted by academics and broadly used by practitioners. The consensus derives from
the model’s advantages, in particular its economic rationality. The DCF method takes
into account the time value of money and has a unique result regardless of investors’
risk preferences (Mun, 2002). In addition, the procedure is clearly defined and can
easily be used by valuers.
Although the DCF method plays a crucial role in valuation, it suffers from at least
three pitfalls. First, the traditional DCF analysis is performed under deterministic
assumptions (for a discussion, see Wofford, 1978; Mollart, 1988; French and Gabrielli,
2004). In other words, one does not take into account uncertainty in the estimated cash
flows; the entire process is therefore devalued when forecasts do not materialise or
even when inputs are slightly manipulated (Kelliher and Mahoney, 2000; Weeks, 2003).
This criticism is particularly severe in real estate valuation since the terminal value,
which is dependent on the last forecasted free cash flow, the perpetual rate of growth
and on the discount rate, is in most cases the largest component of the present value. If
The current issue and full text archive of this journal is available at
www.emeraldinsight.com/1463-578X.htm
The authors thank Se
´verine Cauchie, Philippe Gaud, Etienne Nagy and Agim Xhaja for helpful
suggestions. The usual disclaimer applies.
JPIF
24,2
102
Received May 2005
Accepted September 2005
Journal of Property Investment &
Finance
Vol. 24 No. 2, 2006
pp. 102-122
qEmerald Group Publishing Limited
1463-578X
DOI 10.1108/14635780610655076
such parameters are not determined very rigorously, the estimated value of a property
can be very far off its market value. When the latter value is known, one can also say
that it is easy to set parameters so as to obtain a present value that is close to it.
Another drawback of the DCF method is that there is a circularity problem when
part of the asset is financed by debt. Indeed, the value of the asset is required to
compute the WACC, but the value of the asset is precisely what we are looking for.
Finally, the discount rate is assumed to be constant through time though research has
shown that prices and returns on financial assets are related more to changes in the
required rate of return than to changes in expected cash flows (Fama and French, 1989;
Ferson and Campbell, 1991). To model the time-varying nature of the required rate of
return, Geltner and Mei (1994) and Clayton (1996) use a vector autoregressive
procedure to analyse returns on private real estate. The latter author, for insta nce, finds
that the risk premium on direct unsecuritised commercial real estate varies over time
and is strongly related to general economic conditions.
In this research, we use the Adjusted Present Value (APV) methodology, developed
by Myers (1974), but by adding Monte Carlo simulations. Under some assumptions, the
APV method yields the same results as the widely used DCF technique (Fernandez,
forthcoming), but it solves the circularity problem created by debt financing
(Achour-Fischer, 1999). In addition, with Monte Carlo simulations, which are based on
statistical measures and probability distributions of the variables that enter in the APV
method, we address the uncertainty issue.
With the APV methodology, the discount rate represents the required rate of return
for fully equity-financed properties. Many data analyses have lead us to conclude that
the Capital Asset Pricing Model (CAPM) is in most cases not applicable to estimate this
required rate of return[1]. First, there are usually not sufficient historical data for direct
real estate investments. Second, an appropriate definition of the market portfolio and in
particular of the relative weight of real estate in such portfolio is difficult. Third, the
returns on indirect real estate investments may be poor proxies for direct real estate
returns (Lizieri and Ward, 2000). This problem is exacerbated when one attem pts to
remove the effect of leverage. Further, historical returns may be poor proxies for
expected future returns (Geltner and Miller, 2001). Finally, as mentioned previously,
most such models assume that risk is constant over time.
The contributions of the paper are as follows. First, we address formally the issue of
uncertainty in valuing real estate. This is achieved by using a Monte Carlo approach
within an APV framework. Further, our approach prevents subjective changes of the
values of the parameters used to compute the terminal value, as these are obtained by
clearly defined models or procedures. Finally, we model the discount rate by
considering that it has two components: a risk free interest rate and a risk premium.
We model the interest rate by using the Cox et al. (1985) model. Such model allows us to
assume that the discount rate is not constant through time and that it depends on the
present level of interest rates and their volatility. We suggest an innovative solution to
estimate the risk premium which is assumed to depend on a real estate market
premium and on property specific attributes. The attributes are measured by selected
hedonic attributes which include the quality of location, the age and the quality of
buildings. Hence our method considers that risk is multidimensional and is not only
related to covariance with the market as posited by the CAPM. In that sense it is more
closely related to Arbitrage Pricing Theory (APT).
Monte Carlo
simulations
103
To continue reading
Request your trial