Derivatives Pricing Approaches to Valuation Models: Sensitivity Analysis of Underlying Factors A. E. Baum + C. J. Beardsley * C. W. R. Ward + + Department of Land Management and Development, The University of Reading, Whiteknights, Reading, RG6 6AW, June, 1999
32
Embed
Derivatives Pricing Approaches to Valuation Models: Sensitivity Analysis of Underlying Factors
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Derivatives Pricing Approaches to Valuation Models:
Sensitivity Analysis of Underlying Factors
A. E. Baum+
C. J. Beardsley*
C. W. R. Ward+
+Department of Land Management and Development,The University of Reading,Whiteknights,Reading,RG6 6AW,
June, 1999
*ISMA Centre, University of ReadingWhiteknightsReadingRG6 6AW
Derivatives Pricing Approaches to Valuation Models:
Sensitivity Analysis of Underlying Factors
Abstract
Derivatives pricing techniques are now being applied to real estate
in the US, UK and Australia. However, the UK lease structure offers
a particular opportunity to explore this means of analysis because
of the complex option-like characteristics of the typical
institutional lease.
The standard upward-only rent review introduces the possibility of
splitting the cash flow from the property into separable components.
These are, (1) a simple annuity for any time period of the contract
and (2) the potential uplift(s) in rental tranches at future rent
reviews;
The simple annuity is easily analysed. The variability of future
rent reviews is more usefully understood with reference to
derivatives pricing. This approach introduces a number of new
variables and techniques into real estate investment analysis, the
two most interesting of which are (1) an ability to value stochastic
processes affecting the underlying property, particularly
2
volatility, and (2) the use of risk-neutral valuation techniques to
give correct values for cash flow derivatives which depend on the
underlying property
The range of assumptions that may be made by model builders is wide
and also includes assumptions about the variability and the term
structure of interest rates, plus the existence of tradable assets
which can be used to map the expected rental growth rates into a
risk neutral framework.
The paper uses Monte Carlo simulation techniques that examine how
these factors affect the values produced for standard UK leases. We
will examine the impact of relaxing the upward-only rent review
assumption both as a means of extending the application of the paper
and also to measure the importance of the upward-only rent review vis-
à-vis alternative forms of rental cash flows.
3
Introduction.
The modelling of option pricing stretches back to the beginning of
the 20th century when Bachelier (1900) developed the first valuation
model assuming that stock prices followed arithmetic Brownian
Motion. After an intervening period of neglect, advances were made
by Sprenkle (1964), Boness (1964) and Samuelson (1965) who
recognised that a more applicable model could be derived from
consideration of the geometric Brownian motion with drift approach.
Boness (1964) in particular recognised that the value of a call
option in a stock would be a function of the present values of the
expected share price and the exercise price. Given a market, in
which the share was traded, the present value of the expected share
price was represented by the current share price.
Boness’ formula could be represented by the expression
C = S0 Prob(1) – X Prob(2) e-rt ()
Where
Prob(1) = sensitivity of the Call option price to changes in
the Share price
Prob(2) = Probability of the share price exceeding the Exercise
price at expiration
S0 = the current share price
X = the exercise price
r = the expected return on the share price
t = the time to expiry of the option
4
While this was demonstrably wrong in that it used, as a discount
rate for the option, the return on the share, it could form the
basis of a normative valuation model provided ad hoc adjustments
were made to the discount rate to reflect the higher risk associated
with the option.
A huge step forward was made by Black and Scholes (1973) who
provided the first explicit general equilibrium solution to the
option pricing problem of valuing European call and put options.
They were able to simplify the choice of discount rate by
recognising that a riskless portfolio consisting of shares and
options could be constructed and valued using the risk free rate of
interest. Intuitively since the value of the portfolio could be
estimated, and the value of the share constituent could be observed
in the market, the value of the other constituent – the option could
be derived. This valuation approach based on the assumption of
arbitrage provided an analogous model to the Boness approach but
with the dramatic result of using the risk free rate as the discount
factor. In the Boness formula above, (1), the prob(2) was no longer
the probability of the share price being above the exercise price
but instead became the equivalent probability if the future share
price was modelled in a risk-neutral world. Whilst this is a neat
mathematical transmogrification, it does produce some difficulties
if the users of the model wish to use the model normatively rather
than to estimate the market value implicit in the market prices
observed. For example, the Black and Scholes model will reveal to
investors, the “fair” value of a call option, given today’s share
price. It is more difficult to use the model to value a call option
for an investor who believed that the share price was wrongly valued
by the market and who wished to increase the expected future price
of the share (by, for example, imposing a specific growth rate in
5
the share price). However one has to recognise that, just as
investors use the CAPM in valuing stocks, so they might wish to use
the positive model of Option Pricing to incorporate their own prior
expectations with the information contained in market prices. To
reflect the normative application, we use in this paper two versions
of the underlying model, the first in which the parameters reflect
market prices and a second in which the expected return from the
underlying asset may contain a mis-priced valuation.
Option Pricing using Monte Carlo Simulation.
For some options, the value of the option depends on the past
behaviour of the underlying asset. For example, options can be
traded on the average price of a security during a specific period.
These options are described as being path-dependent and generally,
they are more difficult to value than the conventional put and call
options. In property lease terms, the exercise price may depend on
the value of the rent that is being paid prior to the exercise date.
Whilst solutions can be found for some path dependent options,
property leases can be characterised as consisting of multiple
options. For example, a common lease in UK commercial property
implies that rents should be reviewed upwards every five years. The
rent can only be increased so at the end of each five-year period,
the tenant and landlord have to agree on the appropriate adjustment
to make in the light of market rental movements for comparable
properties. Consider, in 1990, a property in an area that has not
experienced change in property rents since the property was first
let in 1985. The rent for the specific property might not be
increased in which case the next question to answer is whether rents
in 1995 will increase to more than the level originally fixed in
6
1985. Alternatively the rent might be adjusted in 1990 and the
option inherent at 1995 is whether the rent should be increased
above the 1990 level. It will be seen that each subsequent rent
review is a multiple option depending on what level of rent has
already been reached for the specific property in any of the
preceding number of years. One can thus identify in a 25-year lease
ten different options (5, 01; 10, 0; 10, 5; 15, 0; 15, 5; 15, 10; 20,
0; 20, 5; 20, 10; 20, 15). Of these ten options, six have (at the
beginning of the lease) unknown exercise prices in that the rent
that would be paid at the time at which the option could be
exercised would not be known at the beginning of the lease.
For this reason, although it might be possible to derive analytical
solutions to the valuation of the option-like characteristics of the
lease, the simplest approach is to use a Monte Carlo method. This
approach has been widely used in valuing options. Boyle (1976) used
Monte Carlo simulation for the first time to value options;. valuing
European call options on dividend paying stock using 5,000 trials
per estimate for various values of the underlying parameters. In
some circumstances, this technique can be used to derive a pragmatic
formulation of the specific valuation model. Chidambaran and
Figlewski (1995), for example, used Monte Carlo iterations of
100,000 varying the underlying parameters and then derived a quasi-
analytical solution by regressing the consequent valuations on the
parameters varied in the simulation.
1 In this list, the first number denotes the year in which the rent is
reviewed, the second indicates the year in which the existing rent was
fixed. Therefore, 20, 10 implies the option of raising the rent in the 20th
year when the rent has not been altered since it was set in the 10th year.
7
In this paper, we used the Monte Carlo method known as Latin
Hypercube Sampling – another name for this is stratified sampling
without replacement. Using this method, probability distribution is
split into n intervals of equal probability where n is the number of
iterations to be performed on the model
In the first iteration one of these intervals is selected using a
random number
A second random number is then generated to determine where
within that interval the probability function should lie
Then the value of that probability is generated
The process is repeated for the second iteration but the interval
used for the first iteration is marked as used and will not be
selected again
This process is repeated for all of the iterations. Since the
number of iterations is equal to the number of intervals each
interval will only have been sampled once and the distribution
will have been reproduced with predictable uniformity over the
range of probabilities.
Latin Hypercube sampling tends to force convergence of a sampled
distribution in fewer sample. As more iterations are run, the amount
of change in the statistics becomes less and less until the changes
meet a required minimum threshold (in our case the mean and
standard deviation of results were less than 0.5%). In addition, we
use a common random number generator when comparing results to
reduce the variance in the sampling procedure (Law and Kelton,
1991).
Previous Research in Real Estate
8
One of the fundamental principles in Finance Theory is that of Value
Additivity. This principle asserts that if there are two assets A
and B then the value of the portfolio combining A and B must be the
sum of the values of A and B. Indeed this lies at the heart of
modern portfolio theory in indicating that the only type of risk
that investors can seek to be paid for bearing is the systematic
component of risk (that is the risk which cannot be diversified
away). In property leases there are examples of valuation which seem
to be inconsistent with this principle, not least the paper by
Capozza and Sick (1991) who examine the inherent option of
redeveloping a property on the value of existing properties. They
showed that although the cash flow implications of not having a
perpetual income from the lease is only around 1% the removal of the
option to redevelop the property during the lease might be as large
as 20%. In other words, the option to redevelop the land might not
be reflected in the values of the existing rental income flows of
the property unless the right to redevelop the property was “owned”
by the land-owner or land-lord,
In general, under-utilised land can be viewed as a call option
because the holder has the right to erect a new property (subject to
planning or zoning restrictions). Geltner, Riddiough and Stojanovic
(1994) investigated the sensitivity of the redevelopment option to
the supply of available land and found that reducing the available
land uses in an area effectively stimulated development at a much
earlier stage than if alternative land uses were available to the
land owner. Capozza and Sick (1993) investigated the disparity in
value between urban and agricultural land using an option-based
approach. Speculative development is often subject to uncertainty
because of the uncertain timing and cost of any development. It can
be argued that development can be viewed as an Option since the
9
holder of vacant land can effectively exercise an option to develop
the land. Titman (1985), Paddock, Siegal and Smith (1988), Williams
(1991), Capozza and Sick (1993), Quigg (1993,1995), Capozza and Li
(1994) and Yamaguchi et al (1995) have all incorporated option
pricing theory to value such development options. The relationship
between land value, risk in the market for Real Estate and the
optimal development were identified by Williams (1991) in examining
the scale of development and who modelled the rental flow as a
geometric Brownian Motion with drift. Capozza and Li (1994) showed
that intensity interacts in important ways with timing, taxes, and
project values.
Perhaps the seminal article in this literature was Grenadier (1995)
who modelled the pricing of occupancy value of property as a
combination of two processes, first a GBM process that controlled
the occupancy value of rent if no construction was involved,
secondly a process that would control the supply of new property
space if construction did take place. In the elegant paper,
Grenadier demonstrates how the combination of the two processes
gives rise to lease–term structure effects in rents in an analogous
way to that of interest rates. However it should be noted that in
order to derive the solution to the equilibrium rent, Grenadier had
to estimate or assume a number of parameters, including a ratio of
space demand to construction costs that will trigger new supply of
space and the instantaneous correlations of the cost and demand
functions with the market portfolio. Clearly, the estimation of
these parameters presents major problems for investors/researchers
wanting to operationalise the model.
Empirical research in the applicability of option pricing models to
real estate has been limited because of the lack of appropriate
10
data. Quigg (1993) carried out an important study on the
development option value using 2,700 residential land transactions
in Seattle. She valued vacant urban land by accounting for the
option to wait to develop land and found that market prices reflect
an average premium of six percent of land value reflecting the
premium for optimal development. She also found that the standard
deviation of individual real estate assets ranged from 18 to 28
percent per annum.
Since the Black and Scholes paper, there has been a huge literature
of expanding the models of option pricing and its applications to
more exotic products and markets. Specifically there has also been
much development using the Binomial Option pricing model (Cox, Ross
and Rubinstein, 1979) which has been applied to real estate
valuation by Ward, Hendershott and French (1998) to analyse the
different types of rent adjustments observed in real estate leases
in the US, UK and Australiasia. The assumption in that paper was
that the property rental flows were spanned by existing assets in
the capital market and could therefore be valued using geometric
Brownian motion.
Geometric Brownian Motion
In our model, we use a discrete time analog of the Black-Scholes
option pricing model.
Letting:
Po = the price of property now – this is known
T = any future time (measured in years)
11
PT = the price of the property at time T – it is a random variable
and its value is not known until time T
Z = a standard normal variable (with mean = 0 and standard
deviation = 1)
= Mean percentage growth rate of the property (per year)
expressed as a decimal
= Standard deviation of the growth rate of the property
(per year) expressed as a decimal.
Using this notation, the future price of the property may be
modelled as:
PT = Po * exp[(µ - )*T + * Z* ]
Chriss (1997) provides an intuitive description of the process of
geometric Brownian motion, which had its origin in a physical
description of the motion of a heavy particle suspended in a medium
of light particles. The light particles move around rapidly, and as
a matter of course, occasionally randomly crash into the heavy
particle. Each collision slightly displaces the heavy particle; the
direction and magnitude of this displacement is random and
independent from all the other collisions. The nature of this
randomness does not change from collision to collision. In the
language of probability theory, each collision is an independent,
identically distributed random event. The mean and standard
deviation of the large particle’s motion depend on the amount of
time that has passed. If the particle’s behaviour is very volatile
in the short run it will be proportionately volatile in the long
run. Imagine property prices as heavy particles that are jarred
12
around by lighter particles, trades. Each transaction moves the
market price slightly. The conclusions of the model are:
If P is a property with a value of Pt at the current time t, and if
P follows a geometric Brownian motion with mean µ and standard
deviation , then the return on P between now (time t) and a future
time T is normally distributed with:
1. Mean = *(T – t)
2. Standard Deviation = *
The pricing of the property is also referred to as an Itô process,
which has an expected drift rate of (1) above and a variance rate of
(2) above. There are two aspects to note from this model. First, the
standard deviation increases in proportion to the square root of
time. Second, the rate of return is adjusted downwards by a factor
of . What this says is that short run returns alone are not
good predictors of long run returns.
To see this, consider a process with no uncertainty such as the
growth of a bank deposit. At a rate of 10% compounded annually a
deposit with grow from a principal sum of, say, £100, to £110 and
then to £121 and then to £138.1 and so on. However, suppose that
there is a 50/50 chance that at the end of the year the return will
either increase by 5% or fall by 5%. There is now uncertainty with
two possible random ‘jolts’ that could occur at each time step. The
result of this is there is an equal probability of a return of (1 +
0.05) being followed by a return of (1 – 0.05) i.e. 0.9975.
Mathematically as the process unfolds this equates to an adjustment
of (1 + x)*(1 – x) = 1- x2.
13
Heuristically, the stochastic component in returns under the
geometric Brownian motion model is assumed to be normally
distributed with a mean of zero and standard deviation of . If x
is a random variable representing the stochastic component of
Brownian motion then:
Var [x] = E [x2] - E [x]2
But, Var [x] = and since E [x] = 0 the above equation implies
that E [x2] = . And since there is a probability of a positive
return being followed by a negative return the average amount the
drift rate will be depressed is because x2 itself represents the
result of two price movements.
In modelling property prices, the effect of geometric Brownian
motion is to produce a lognormal probability distribution for the
value of the property. Such a distribution for year 10 of our
normative model (described below) is shown:
14
Parameters in Option Pricing Model
The Black and Scholes (1973) model of option pricing requires the
following parameters; current asset price, exercise price, risk-free
rate of return, the time to expiry of the option and the volatility
of returns of the underlying asset. In the case of income paying
assets, the model can be modified to reflect either the present
value of the income received from owning the asset or the equivalent
continuous income yield of the asset. In either case, the value of
the option is reduced, as the capital growth of the underlying asset
is lower, the higher the income received by owners of the underlying
asset.
The crucial assumption in using the Black and Scholes model is that
the asset is spanned by existing market assets which can be used to
form an instantaneous hedge portfolio which is riskless in the short
term. This then permits the assumption of the asset price being
modelled by geometric Brownian motion (GBM) and the substitution of
the riskless rate of interest for the risk-adjusted return on the
asset with a great advantage of computational simplicity of the
model valuations. In the Boness model, referred to in the
introductory paragraphs of this paper, the use of the ‘wrong’
discount rate tended to produce valuations of the option that were
too high. However if the assumption of spanning markets is not
justified then the use of GBM is formally flawed and the consequent
valuation of the option-like characteristics of the property lease
is also only an approximation of the equivalent market prices.
15
As Dixit and Pindyck (1994) explain, the relaxation of the
assumption of spanning implies that the expected return on the
underlying asset is no longer the risk free rate but an adjusted
return. Grenadier echoes this discussion in his model,
“…In the case of leasing markets, where theunderlying assets such as office buildings aresubject to substantial transactions costs,indivisibility, and the inability to be soldshort, such arbitrage arguments areparticularly questionable.”
(Grenadier, 1995, p304)
In these conditions, the expected return on the asset is reduced to
a level that is equivalent to the risk neutral return by subtracting
from the expected return a premium which reflects the systematic
risk of the non-tradable asset.
Or
( )
In this paper, we first use the heroic assumption that a tradable
property asset does exist and then we relax this assumption and use
a process as defined by (2) above to investigate how sensitive the
valuation is to changes in the assumptions of the model. This
methodology corresponds to factoring in the “market price of risk”
or Lambda discussed in Hull (1997) and Shimko (1992)
The UK Lease
16
In this paper, we are analysing the conventional UK institutional
lease that is specified for a maximum of 25 years with rent being
reviewed every five years. The rent is changed only upwards so if
market rents have fallen over a period of time, the tenant is left
paying a higher rent than available on other properties more
recently let. In practice there may be some concessions to tenants
such as break clauses but we ignore this issue in this paper. We
also ignore any reversionary value to the landlord at the end of the
lease since this contribution to the value of the property would be
the same regardless of the terms of the 25-year lease.
Following Grenadier, we assume that the asset value (which in our
initial case is assumed to be traded) is a perpetuity that offers a
constant income yield. The asset price is assumed to follow GBM. The
income from such an asset is then used to benchmark the rental value
of the individual property. Since we are not including the
reversionary value the lease can be thought of as the present value
of the annual rental flows. Thus in a risk neutral world the rental
flows are growing at a rate which is negatively related to the
initial yield of the asset. In terms of an equity investment this is
equivalent to saying that the dividend growth is inversely related
to the dividend yield.
We then create a lease which is based on the annual rent but fixed
for period of five years and then adjusted either to the higher
figure of the annual rent or remaining at the level attained at the
current point in time. Of course, we recognise that this will be
less valuable in present value terms than the annual rental flows.
We could have adjusted the five year rent to provide an equal
present value to the annual revised rent, but it is simpler to
17
represent the differences in terms of the total value of the lease
rather than an adjusted rent to provide the same values.
We arbitrarily fix the risk free rate at 5% and an initial yield of
4%. In terms of the five-year review, this provides an implicit
growth rate of 1.02% to provide the equivalent total return in a
risk neutral world.
Volatility
Property is usually characterised as a low volatility asset;
quarterly rents and yields often appear to move “stickily” in
response to changes in the economic environment. However over longer
horizons it is apparent that investors and developers appear
surprised at changes in rents and returns. This would suggest that
the underlying volatility is greater than that indicated by
quarterly returns. This suggestion is confirmed by estimates from
Lee and Ward (1999) who find, using a decomposition of the returns
from different portfolios, estimates of annual volatilities of
individual properties ranging from 16% to 32% and an average
correlation between properties of just over 0.15. This is a return
volatility which corresponds to the modelling of the asset price
returns used by Grenadier (op.cit,)
The volatility of returns was therefore specified to be a minimum of
zero (to correspond with the conventional DCF approach to valuing UK
property) and a maximum of 40%. The asset prices are then simulated
10,000 times using a common random number to provide different
rental paths, which are then summarised in the following table.
Table 1: The Effect of volatility on Lease ValueAll values expressed in £ million
18
Volatility 0% 10% 20% 30% 40%Value of Upwards-
only lease
60.
3
64.
9
71.
8
78.
9
86.
0Value of Annual
reviewed lease
62.
3
63.
3
62.
2
62.
1
61.
9Ratio of Upwards-
only to freely
floating
0.9
7
1.0
4
1.1
5
1.2
7
1.3
9
As expected, the effect of volatility is to increase the value of
option-like lease. In the case of zero volatility, the upwards-only
lease is less valuable than the annual review because the initial
rents are fixed at the same level. Clearly if there is any implied
rental growth, the step-pattern of the five year review pattern will
be less valuable than the annual review.
The normative case is shown in Table 2. In this case, the investor
believes that the parameters are misvalued and expects that the
property itself will outperform the rates implied by the market. In
the example used, we assumed that the rental growth rate was 2% more
than the rate implied by the present market prices. Whilst this
steps beyond the intrinsic arguments of the arbitrage based option
pricing model, it is a pragmatic attempt to deal with the present
market structure in the UK where the present yields in the property
market would appear to be so far out of line with the yields
available from bonds and equities that property would seem to imply
a negative rental growth or a huge risk premium.
Table 2: The Effect of volatility on Lease ValueThe Normative case – Abnormal rental growth of 2%
All values expressed in £ million
Volatility 0% 10% 20% 30% 40%
19
Value of Upwards-
only lease
71.
5
74.
0
81.
0
89.
2
97.
6Value of Annual
reviewed lease
76.
9
77.
0
77.
1
77.
3
77.
4Ratio of Upwards-
only to freely
floating
0.9
3
0.9
6
1.0
5
1.1
5
1.2
6
In this case, as expected the option effects are less important
because the rental growth effect dominates the implied rental growth
rates in the positive model. The implication is that if property is
undervalued the premium values that might be backed out of market
prices are too high. Of course if it were really possible to
construct an arbitrage portfolio, the difference between the two
tables would disappear; but in the absence of a practical arbitrage
portfolio, the differences remind us that the assumption of tradable
asset is an important element of option pricing when applied in the
real world.
Interest Rates
In Tables 1 and 2, we assume that the riskless rate of interest in
known and constant but in practice we observe a term structure of
interest rates that imply changing discount rates over time. In this
paper we look at the sensitivity of the values to changes in the
modelling process of interest rates. We relax the assumption of
constant interest rates and instead assume a model based on Cox,
Ingersoll and Ross (1985) in which the yield curve converges from
the short-term rate to a longer term equilibrium rate.
20
The Cox, Ingersoll and Ross model specifies the mean reverting
process as the following:
where a, b, and are constants. The short rate is pulled to a level
b at rate a. Superimposed on this “pull” is a normally distributed
stochastic term that has a standard deviation proportional to .
This means that as the short term interest rate increases, its
standard deviation increases.
Since the present yield curve is downward sloping, we assume that
the yield converges to 4% with different variability. The volatility
of the underlying asset is kept constant at 20%. The parameter, a,
is kept constant at 6%.
Table 3: The Effect of Mean Reverting Interest RatesAll values expressed in £ million
Volatility 20%
no
yield
reversi
on
20%
Std.dev.
of yld=
0.5%
20%
Std.dev. of
yld= 1.0%
Value of
Upwards-only
71.8 81.2 81.3
21
lease Value of
Annual
reviewed lease
62.2 64.8 64.9
Ratio of
Upwards-only
to freely
floating
1.15 1.31 1.31
As the imposition of the downward sloping yield curve reduces the
average interest rates, the introduction of mean reverting interest
rates increases the value of the lease as the discount rate is
reduced. It will be seen that the effect of increasing the
variability of the interest rate shifts has little or no effect on
the valuations. This result suggests that the assumption of mean –
reversion in the yield structure can probably be adequately captured
by reducing (or increasing) the average rate of interest rather than
using a stochastic interest rate model.
Underlying Asset: Tradable vs. Non-Tradable
It is in this area that there is most scope for error. As explained
above, the assumption of tradable assets allows the researcher to
apply a risk-neutral rate of interest as the total return on the
underlying asset. Without this assumption, the total return on the
asset can diverge from the risk neutral return. Given the
characteristics of property, the adjustment is, to a large extent,
arbitrary, influenced both by conceptual and empirical decisions.
In the equation (2) above, the total return has to be adjusted by
subtracting from the expected market return, the product of the risk
22
market premium and the individual systematic risk. In this paper, we
take the FTA All Share Index as representing the traded asset on
which the property depends. To capture the systematic risk of the
property, we regress the IPD Annual Property Index (returns) against
the All Share Index (returns). However because the Property Index is
an appraisal-based index, it is treated as an illiquid asset and
the regression is run following the Dimson method, i.e.
(3)
Table 4: Estimated Beta (Dimson) of PropertyWith respect to Equities
Adjusted R-squared = 0.15.In the regression, we used contemporaneous returns and two lagged returns in the regression on the assumption that the stockmarket ‘leads’ the property market.
Using this methodology, the estimated beta was 0.4, the mean return
on the Stock market was
21% with a standard deviation of 25% and the risk free rate was, as
before 5%. The results are shown in Table 5. These results could be
23
suspect because the historical risk premium of the equity market is
very high if measured over the 1971-1998 period. An alternative
result could be found if one used the risk premium of the equity
market over the longer period 1919-1978. This widely quoted premium
of 8% (see Brealey and Myers, 1998) is often used as an estimate of
the forward-looking risk premium on equities. This is incorporated
in the analysis and the results are reported in Table 6. As can be
seen, the results are sensitive to the equity premium chosen. Since
there is no theoretical or a priori justification for a particular risk
premium, this must be seen as a major issue when the option-pricing
model is applied to property.
Table 5: The Effect of Non-tradable AssetsNormative case assumptions above
Historical Risk Premium on Equity Market: 16.5%, 1971-1998
All values expressed in £ million
Volatility 20%
= 0
20%
= 0.5
20%
= 1Value of
Annual
reviewed lease
77.1 44.4 29.4
Value of
Upwards-only
lease
81 63.3 60
Ratio of 1.05 1.43 1.97
24
Upwards-only
to freely
floating
Table 6: The Effect of Non-tradable AssetsNormative case assumptions above
Historical Risk Premium on Equity Market: 8%, 1919-1978
All values expressed in £ million
Volatility 20%
= 0
20%
= 0.5
20%
= 1Value of
Annual
reviewed lease
77.1 65.8 56.7
Value of
Upwards-only
lease
81 75 69.6
Ratio of
Upwards-only
to freely
floating
1.05 1.14 1.23
However, one need not stop at the choice of one risk premium. In
principle, the traded portfolio could consist of any relevant
portfolio and an obvious case could be made for the choice of bonds.
We therefore estimated the Dimson beta for property against the
Long-term government bond index reported in the IPD Annual Report
1999. This is presented in Table 7.
25
Table 7: Estimated Beta (Dimson) of PropertyWith respect to Bonds
Source: IPD Annual Index Report (1999)Adjusted R-squared = 0.08.In the regression, we used contemporaneous returns and two lagged returns in the regression on the assumption that the bondmarket ‘leads’ the property market
As can be seen, the property index has a higher Beta with respect to
the bond index than it does with equities. (The statistical
insignificance of the estimates is not an issue in the Dimson
methodology since one is only looking for unbiased parameter
estimates). However, it was found that the risk premium for bonds
was 8.5% over the 1971-1998 period, a figure that is quite close to
the equity risk premium for the 1919-1978 period. In other words,
the results from varying the market traded asset from equities to
bonds resulted in a higher systematic risk factor but lower risk
premium. Therefore the results are consistent with the case already
reported in Tables 6 and 7; the important variable appears to be the
market risk premium.
26
Conclusions
The purpose of this paper was to explore the sensitivity of the
Option pricing model lease valuation to changes in the parameters
commonly used. The importance of volatility is readily understood
but the more crucial aspect is shown the estimated risk premium on
the traded asset if the assumption of property leases being traded
is relaxed.
In summarising the results of this investigation, it is convenient
to refer to Table 8 which presents the differing estimates of the
premium on a upwards-only rent review under different assumptions.